Physical AI Brief

Forecasting the evolution of dynamic environments is crucial for autonomous agents. While generative world models have recently achieved high photorealism in 2D video synthesis by mixing ego-motion and environmental dynamics within the image plane, they exhibit physical inconsistencies, such as morphing or vanishing objects, especially over long time horizons. In this paper, we propose FR3D, a world model that predicts a persistent 3D latent representation for future dynamic 3D reconstruction. Unlike prior works that treat the world as a sequence of image-based features, FR3D explicitly decouples the 3D evolution of the scene from the agent's trajectory, treating the inferred ego-motion as a latent proxy for action. This disentanglement resolves the ambiguities between self-motion and world-motion, ensuring geometric consistency into the future. Furthermore, we introduce a teacher-student distillation strategy that leverages the spatial "common sense" of off-the-shelf foundation models, leading to robust zero-shot generalization. Extensive experiments demonstrate FR3D's strong performance for future dynamic 3D reconstruction from monocular observations across multiple datasets, even 2 seconds into the future. Project page: https://fr3d-wm.github.io.

Unified Multimodal Modeling aims to integrate visual understanding and generation within a single system. However, existing approaches typically rely on two disparate visual tokenizers, which splits the representation space and hinders truly unified modeling. We propose UniAR, a unified autoregressive framework where a single discrete visual tokenizer serves as the key bridge between understanding and generation, enabling a shared context in which the model can directly interpret its own generated visual tokens without additional re-encoding. UniAR adapts a pretrained vision encoder with multi-level feature fusion and a lookup-free bitwise quantization scheme, preserving both high-level semantics and low-level details while scaling the effective visual vocabulary at minimal cost. Building on this, the unified autoregressive model adopts parallel-bitwise-prediction to jointly predict spatially grouped, multi-level visual codes, substantially reducing visual sequence length and accelerating generation. Finally, a diffusion-based visual decoder operates on discrete visual tokens to decode high-fidelity images. Through large-scale pre-training, followed by supervised fine-tuning and reinforcement learning, UniAR achieves state-of-the-art performance on image generation and image editing while remaining competitive on multimodal understanding benchmarks. The project page is available at https://sharelab-sii.github.io/uniar-web.

Robots deployed in the real world should learn from their experience and improve over time. This requires a mechanism of practicing and learning from feedback. In this paper, we propose VERITAS, a generator-verifier framework for generalist robot policies for inference-time policy steering and self-improvement. We use a pre-trained generalist robot policy as a ``generator'' and pair it with a gradient-free ``visual verifier'' that evaluates actions at inference time. This framework enables inference-time steering that improves policy performance without additional training. We demonstrate that inference-time verification consistently outperforms vanilla generalists without training on additional demonstration data. Additionally, we demonstrate that the verified rollouts provide effective supervision for offline policy improvement: policies fine-tuned on verified self-generated trajectories achieve consistent performance gains. Notably, we find that post-training with verified rollouts achieves comparable efficiency to expert demonstrations, while requiring no human interventions. Our results highlight inference-time verification as a practical and scalable mechanism for improving robotic policies during deployment.

Scaling model size, specifically depth and width, has driven significant progress in transformer-based language models. However, most architectures maintain a constant width across all layers, allocating a fixed parameter and computation budget evenly despite different layers potentially playing distinct computational roles. In this work, we empirically investigate nonuniform capacity allocation across network depth by proposing a $\times$-shaped > <former architecture. This design maintains wider early and late layers while narrowing the middle layers, utilizing a parameter-free residual resizing mechanism. Across decoder-only language models ranging from 200M to 2B parameters (dense) and 3B parameters (MoE), our > <former consistently outperforms parameter-matched uniform baselines on language modeling loss. By reducing the average layer width, this architecture also requires fewer overall FLOPs (22% reduction under fitted loss-matched scaling curves) and smaller KV cache memory and I/O cost (15% reduction). In analysis, we show that this bottleneck structure results in qualitatively different representations in residual streams. Overall, our results demonstrate that nonuniform width allocation can result in more resource-optimal scaling of language models.

Collaborative human-object interaction shows dynamic and complex movements that require mutual anticipation and continuous adjustment between participants and the shared object. Modeling such collaborative multi-human object interaction (MHOI) scenarios requires high-quality data acquisition as a foundational step; however, this is challenging due to the inherent complexity of MHOI where human-human and human-object interactions occur simultaneously. Such complexity leads to noisy MHOI captures characterized by several artifacts: contact misalignment between hands and objects, motion jitter and temporal inconsistencies in the captured sequences, and missing or incomplete finger-level articulation details. To address these challenges, we present MOCHI (MOtion Enhancement of Collaborative Human-object Interactions), a two-stage framework for enhancing noisy MHOI data. Our approach first generates physically plausible hand grasps through optimization from noisy body input, producing grasps that are both physically plausible and semantically consistent with the body pose, where these optimized grasps are extended into complete hand-object interaction sequences. Consequently, the full-body motion for all participants are refined through a diffusion-based noise optimization framework that uses single-person motion priors. During the optimization process, we introduce optimization objectives to encode human-object and human-human interaction information within these single-person priors. Experimental results demonstrate the effectiveness of our pipeline across diverse MHOI data, either acquired by existing capture methods or synthesized by generative models. We further show robustness of our system across varying numbers of participants and types of interactions, and demonstrate various applications including keyframe-based MHOI creation and data augmentation through varying object geometries.

Event cameras sense the world through asynchronous brightness changes with microsecond latency and high dynamic range, offering motion fidelity far beyond frame-based sensors and capturing temporal structure that conventional exposures often miss. These properties make events a powerful complement to RGB in autonomous driving, especially under blur, glare, and rapid motion, where frame-based perception can become unreliable. However, existing event-aware vision-language models remain limited to generic perception and do not reveal how event sensing contributes to reasoning and decision-making across the full driving loop. We present EventDrive, a large-scale benchmark and model suite that unifies event streams, RGB frames, and language supervision across four core dimensions: Perception, Understanding, Prediction, and Planning, covering captions, structured QA, grounding, motion-state recognition, trajectory forecasting, and planning tasks. Building on this foundation, EventDrive-VLM introduces a multi-horizon event pyramid and a temporal-horizon mixture-of-experts module to adaptively encode and fuse asynchronous and frame-based information for downstream reasoning. Comprehensive evaluation across diverse tasks shows that event streams provide substantial gains in temporal precision, motion awareness, and robustness, bringing event sensing into the center of driving intelligence.

We present EBench, a simulation benchmark that diagnoses generalist mobile manipulation policies beyond a single success-rate scalar. EBench comprises 26 diverse and challenging manipulation tasks annotated along 5 capability dimensions and 4 generalization dimensions. We evaluate state-of-the-art generalist manipulation models including $π_0$, $π_{0.5}$, XVLA, and InternVLA-A1, and reveal that models with near success rates exhibit strikingly different capability profiles: $π_{0.5}$ achieves the highest test success rate and the best train--test retention, whereas InternVLA-A1 dominates mobile manipulation but collapses on dexterous tasks, and XVLA exhibits strengths on a disjoint set of atomic skills compared to other policies. Beyond capability profiling, EBench analyzes the generalization ability from 4 representative perspectives, identifying the impact of different distribution shift factors. The results reveal strengths and weaknesses of models behind an overall score. We hope this benchmark offers a broad set of diagnostic signals to guide iteration on generalist manipulation models.

Reproducing research results from papers and released code is central to scientific progress. Existing works have introduced benchmarks to evaluate whether LLM agents can assist with reproducibility, but they are difficult to scale due to their reliance on substantial manual effort for data curation and evaluation. We introduce ReproRepo, a scalable framework for reproducibility evaluation that leverages human-raised GitHub issues as naturally occurring supervision on realistic reproduction blockers. We instantiate ReproRepo on 1,149 recent machine learning papers from major conferences and evaluate four frontier model-agent configurations. Our results show that LLM agents, even without executing code, can identify many real-world reproducibility problems from paper-repository pairs: the best agent in our study, namely Codex with GPT-5.5, surfaces at least one semantically related human-reported blocker for ~90% of papers in the study. Further analysis shows that agents are particularly effective for surfacing visible failures and identifying the right semantic region, but may still be insufficient in exact localization. ReproRepo can serve as a reusable, scalable framework for future evaluations of LLM agents on real-world reproducibility auditing. Our code is released at https://github.com/LithiumDA/ReproRepo.

Zero-Shot Object-Goal Navigation (ZS-OGN) requires embodied agents to explore and locate target objects without any prior training. To this end, recent methods leverage foundation models. But they typically rely on static priors and lack adaptation, which leads to repeated errors and costly trial and error. In this paper, we propose a self-evolving ZS-OGN framework that enables continuous test-time improvement. Specifically, we build an agentic rule memory by extracting actionable knowledge from past trajectories. Then, we propose a retrieval strategy based on upper confidence bound, selecting effective rules by balancing semantic relevance and historical success. In addition, we introduce a memory-guided preflection module that forecasts potential outcomes before action, reducing inefficient exploration. Extensive experiments show that our method outperforms existing zero-shot baselines, achieving a 10.1\% improvement in success rate with fewer unnecessary steps.

Accurate mechanical properties (or materials) Young's modulus ($E$), Poisson's ratio ($ν$) and density ($ρ$) are essential for reliable physics simulation of digital worlds, but most 3D assets lack this information. We propose AdaVoMP, a method for predicting accurate dense spatially-varying ($E$, $ν$, $ρ$) for input 3D objects across representations, improving the resolution, accuracy, and memory efficiency over the state-of-the-art. The foundation of our technique is a sparse and adaptive voxel structure SAV that efficiently represents both the input 3D shape and the material field output. We replace the fixed-voxel model of the most accurate prior method, VoMP, with a novel sparse transformer encoder-decoder model that learns to generate a unique SAV autoregressively for every input shape to represent its materials, achieving a resolution $16^3\times$ higher than prior art. Experiments show that AdaVoMP estimates more accurate volumetric properties, even with lesser test-time compute than all prior art. This allows us to convert high-resolution complex 3D objects into simulation-ready assets, resulting in realistic deformable simulations.

With sophisticated cyber-attacks becoming increasingly prevalent, modern networks require intelligent autonomous cyber-defense agents trained via Reinforcement Learning (RL). These agents employ neurosymbolic approaches such as behavior trees with learning-enabled components (LECs) to learn, reason, adapt, and implement security rules while maintaining critical operations. However, these autonomous networks are partially observable systems, i.e., the cyber-attacker's (red agent's) actions are not observable, making it difficult for the defender to predict red actions, learn red policies, or assess the attacker's intrusion levels. To address this, we propose a Policy Learning Technique using imitation learning to learn policies for partially observable RL agents with discrete states and discrete actions. We apply this technique in an autonomous cyber environment to predict red agent's actions from network observations and defender actions. Integrated with a neurosymbolic cyber-defense agent, our method effectively handles different red policies and achieves high prediction accuracy across diverse simulated scenarios.

Knowledge distillation transfers a teacher's competence to a small student but is brittle in the small-student regime: forcing the student to imitate logits from a much larger teacher concentrates it on the teacher's sharpest modes, hurting generalization on benchmark families beyond the training corpus. Reinforcement learning (RL) avoids logit imitation by training on the student's own rollouts. However, on questions where every rollout fails-yielding zero advantage and being silently discarded-injecting a stronger teacher's response into the policy gradient breaks the on-policy assumption and induces drift. We introduce Zone of Proximal Policy Optimization (ZPPO), inspired by Vygotsky's zone of proximal development, which keeps the teacher inside the prompt rather than the policy gradient. On hard questions, ZPPO constructs two reformulated prompts: a Binary Candidate-included Question (BCQ) pairs one correct teacher response with one incorrect student response as anonymized candidates the student must discriminate, and a Negative Candidate-included Question (NCQ) aggregates the student's wrong rollouts into a single prompt to surface their shared failure modes. A prompt replay buffer recirculates each hard question until it either graduates-the student's mean rollout accuracy on it reaches half- or is FIFO-evicted under finite capacity, amplifying BCQ and NCQ inside the student's current zone of proximal development. On the Qwen3.5 family at four student scales (0.8B-9B) with a 27B teacher, post-trained as vision-language models and evaluated on a 31-benchmark suite (16 VLM, 10 LLM, 5 Video), ZPPO outperforms off/on-policy distillation and GRPO, with the largest gains at the smallest scale.

Dataset distillation (DD) has emerged as a prominent approach in data centric machine learning, aiming to synthesize compact training sets for efficient training by compressing the information in large datasets into a small number of synthetic samples. However, DD methods are often evaluated under inconsistent evaluation protocols, ranging from standard ERM to single/multi-teacher supervision, making it difficult to isolate the effectiveness of distilled data from evaluation. Moreover, many prior methods claim that DD outperforms data pruning approaches such as coreset selection (CS), based on the assumption that restricting condensed datasets to subsets of real samples fundamentally limits their expressiveness. In this work, we critically evaluate DD methods through large-scale experiments using standardized datasets and evaluation protocols to assess their intrinsic effectiveness. We benchmark seven state-of-the-art (SOTA) DD methods on ImageNet-1K, ImageNet100, and ImageNette, using three widely adopted training protocols against three CS strategies. Our results show that while some DD methods fail to outperform even simple random subsets, the SOTA DD approaches are comparable to or worse than coresets on large-scale datasets and incur a substantially higher cost for construction. Beyond accuracy, we also evaluate the representativeness, diversity, and quality of condensed sets, and find that coresets consistently achieve better coverage of the original data distribution. These findings highlight the limited practical advantages of current DD methods and show that coresets remain competitive and are often a more computationally efficient alternative for data-centric learning.

Current world models face a fundamental tension: faithful long-horizon simulation demands deep computation, but deeper models are expensive to deploy and prone to compounding errors. We resolve this by introducing Looped World Models (LoopWM), which are the first looped architectures for world modelling. Our method iteratively refines latent environment states through a parameter-shared transformer block. This yield up to 100x parameter efficiency over conventional approaches with adaptive computation that automatically scales depth to match the complexity of each prediction step. Orthogonal to scaling model size and training data, LoopWM establishes iterative latent depth as a new scaling axis for world simulation, which might significantly push the community forward.

Looped architectures provide an inductive bias toward learning step-by-step procedures for tasks that require compositional reasoning. The number of effective layers reached by looping determines the quality of the solution these models find. Like deep architectures, looped architectures are prone to a signal propagation problem induced by depth as the halting decision is postponed. In this paper, we address this signal propagation issue using pre-norm layers and residual scaling. Building on these architectural modifications, we propose FPRM, a Transformer-based Fixed-Point Reasoning Model that uses fixed-point convergence as an end-to-end halting mechanism in a looped architecture. We show that fixed-point halting allows FPRM to adapt its compute to task difficulty. FPRM is effective on common reasoning benchmarks, namely Sudoku, Maze, state-tracking, and ARC-AGI.

The LLM-empowered personal health agents with user health (sensor) metrics have offered a promising pathway to alleviate global disparities in healthcare access. However, large-scale clinical deployment remains constrained by an open-ended evaluation bottleneck: physician annotation is reliable but costly and unscalable, while LLM-as-a-judge evaluators are scalable but subjective, inconsistent, and sometimes clinically misaligned. We introduce RubricsTree, a scalable evaluation framework with an expert-aligned hierarchical taxonomy of over 100 atomic, clinically-verifiable Boolean rubrics, evolving from the insights of 4,000 real user queries through an iterative human-in-the-loop curation protocol with an expertise panel led by an experienced physician. A context-aware adaptive router activates only the relevant auto-weighted rubric subset per query, providing the throughput needed for scalable evaluation with expert-aligned quality. Through a systematic meta-evaluation, we show that RubricsTree (i) substantially exceeds a strong large-scale evaluation baseline in expert alignment on challenging open-ended queries; (ii) reliably penalizes contextually degraded responses; and (iii) when used as structured instructions, text feedback, or training rewards for performance optimization, yields up to ~66% relative gains on HealthBench for Gemini, GPT, and Qwen model families. RubricsTree thus provides a scalable, auditable, and evolving evaluation infrastructure required for the continuous optimization of product-level personal healthcare AI.

Agent skills are emerging as an important attack surface in LLM-based systems. Through an empirical study of existing skill scanners, we find that current defenses primarily rely on textual descriptions, manifests, and source code as the main signals for security analysis, which can leave visually conveyed malicious intent insufficiently examined. This creates a practical blind spot: harmful operational instructions hidden in images may bypass scanning while still being recoverable by multimodal agents during deployment. To systematically investigate this threat, we propose SkillCamo, a document-mediated multimodal instruction attack that conceals malicious instructions within images bundled with a skill while rewriting the surrounding documentation to naturally reference those images as part of the normal workflow. Thus, the attack does not rely on the image alone, but on the joint interpretation of textual guidance and visual payload at execution time. To defend against such attacks, we further propose ExecScan, an execution-grounded multimodal scanning module that performs intent extraction, behavior reconstruction, abuse assessment, and deliberative execution simulation over skill artifacts. ExecScan jointly analyzes documentation, code, referenced resources, and visual content to recover hidden instructions, reconstruct executable behavior chains, and identify downstream risks such as exfiltration, destruction, persistence, deception, and privilege escalation. Extensive experiments show that image-hidden malicious instructions challenge existing skill scanners, while ExecScan can improve the skill scanning performance.

On-policy self-distillation (OPSD) has proven effective for post-training large language models (LLMs), yet its application to diffusion LLMs (dLLMs) remains unexplored. Existing OPSD methods are inherently autoregressive-centric. They inject privileged information via left-to-right prefix conditioning with token-level divergence supervision, a design that fundamentally conflicts with the arbitraryorder generation of dLLMs. We introduce d-OPSD, the first OPSD framework tailored for dLLMs. Our approach makes two core contributions. First, we reframe self-teacher construction by using self-generated answers as suffix conditioning, enabling the student model to learn from "self future-experience" rather than privileged prefixes. Second, we shift supervision from token-level to step-level, aligning training with the iterative denoising process of dLLMs. Experiments across four reasoning benchmarks show that d-OPSD consistently outperforms RLVR and SFT baselines with superior sample efficiency, requiring only around 10% of the optimization steps by RLVR and opening a promising pathway for dLLM posttraining. The code is available at https://github.com/xingzhejun/d-OPSD.

We evaluate the adversarial robustness of two frontier large language models (LLMs) developed by Anthropic, Fable 5 and Opus 4.8, against four families of automated jailbreak attack across 7 826 harmful intents spanning a ten-category harm taxonomy. Using the HackAgent red-teaming framework, hundreds of thousands of adversarial attempts were generated and every apparent success was independently re-adjudicated by a panel of three judge models (majority vote). Both models resist the majority of attacks, but the residual surface is larger than aggregate framing suggests: it is dominated by adaptive iterative attacks, while static obfuscation is near-fully neutralised. The strongest adaptive search (tree-of-attacks) breaks Opus 4.8 on 11.5% of intents overall, whereas Fable 5 stays in the single digits (6.1% worst-case). Aggregate rates therefore should not be read as reassurance. Even in these hardened configurations, the two models produced 1 620 (Opus 4.8) and 702 (Fable 5) panel-confirmed harmful completions spanning every harm category, located automatically, cheaply, and within the first one or two refinement steps by an attacker model with no human expert in the loop. The reasonable conclusion is that even the best, most-tested frontier models remain reliably breakable under sustained automated pressure.

As high-quality public web corpora become increasingly exhausted, clean long-context documents have become a scarce and expensive source of training data for large language models (LLMs). Existing long-context corpora are often proprietary and costly to acquire, synthetically generated, or concentrated in narrow domains such as programming. We introduce the Stanford EDGAR Filings Dataset (SEFD), an open reconstruction of SEC filings into layout-faithful MultiMarkdown for financial language modeling and evaluation. SEFD makes audited financial statements, risk disclosures, ownership reports, accounting notes, and market-moving event filings usable as long-context pretraining data and as a basis for financial reasoning, forecasting, compliance, and document understanding. The resulting corpus is token-efficient, model-ready, and has less than 0.1% overlap with Common Crawl-derived corpora. We release SEFD-v1, a 152B-token initial public snapshot, and provide corpus-level analyses of a larger 18.5M-filing archive estimated at 550B tokens. We further introduce two SEFD-derived benchmarks: EDGAR-Forecast, which evaluates filing-grounded numerical forecasting after model knowledge cutoffs, and EDGAR-OCR, which evaluates transcription of complex financial tables.

Deep research (DR) systems are increasingly used for complex information-seeking tasks, but existing works mainly focus on generating reports and summaries. In contrast, many enterprise tasks instead require an agent to identify concrete workflows which is a sequence of action-steps. For example, rather than summarizing budgeting policies, an agent should be able to determine the steps needed to answer a question such as: "How do I request new headcount given a fixed budget?". Therefore, we introduce DRFLOW, a benchmark for evaluating personalized workflows predicted by agents from heterogeneous sources. Each task requires the agent to identify relevant evidence from scattered sources, then use that evidence to predict the correct action-step sequence for the user's task. DRFLOW contains 100 tasks across five domains, with 1,246 reference workflow steps grounded in more than 3,900 sources. We define seven diagnostic metrics covering factual grounding, step recovery, structural ordering, condition resolution, and personalization. We further present DRFLOW-Agent (DRFA), a workflow-oriented reference agent to predict personalized workflow. We show that although DRFA improves over strong baseline agents (upto 10.02% average F1 score), there is substantial room for improvement remains across these workflow metrics, indicating that predicting complete and correct personalized workflows remains a challenging frontier for deep research.

Conventional human-in-the-loop approaches typically involve users only when a robot encounters failure or uncertainty, treating humans primarily as tools for improving robot performance. However, in many human-centered robotics settings, interaction should support engagement by keeping users involved in decision-making rather than limiting them to failure-driven interventions. This is particularly compelling in physical caregiving, where mobility limitations can reduce users' ability to intervene or modulate the robot's behavior in the moment. As a result, failure-driven interaction policies may relegate users to passive observers for long stretches of the task. For example, a user with mobility limitations may feel less engaged when being continuously and passively fed by a robot. At the same time, overly frequent interaction can be tiring and increase the user's workload. To address this trade-off, we propose Engagement-aware MPC (E-MPC), a user-engagement-aware method that plans interaction to maintain engagement while respecting a workload constraint. E-MPC leverages a user interaction dynamics model that captures how user engagement evolves as a function of both the frequency and type of interaction. Rather than requesting input only when difficulties arise during task execution, the robot proactively considers the user's preferred level of engagement throughout the task, balancing autonomy and interaction while ensuring task success. We evaluate E-MPC in simulation with several ablations and baseline comparisons. Results demonstrate the effectiveness of our approach across diverse user personas. In addition, we conduct a real-world user study with participants with emulated mobility limitations on a robot-assisted bite acquisition system, showing that E-MPC improves user experience while maintaining task success.

Finite-dimensional (FD) diffusion policies exhibit temporal drift owing to discretization artifacts that degrade long-horizon performance (when deployed on physical systems). We introduce a backward Kolmogorov equation that lifts diffusion policies to a Cameron-Martin space -- a subset of the Hilbert space. Essentially, replacing stochastic score matching with a deterministic boundary-value PDE problem. Our core innovation thrives on Gaussian measure theory whereupon the diffusion noise covariance operator is realized from a colored noise distribution which prescribes a notion of regularity on samples from the model at inference time. We train the diffusion model with a derived precision-weighted Cameron- Martin loss and a Kolmogorov residual is introduced as a PDE diagnostic during inference. These substitutions yield (i) convergence guarantees where the bound's constants depend on the effective rank of the kernel rather than action dimension, (ii) improved trajectory regularity via spectral weighting, and (iii) a deterministic failure detector without reward signals. Validation across two application domains demonstrates substantial improvements: on the PushT manipulation benchmark, the Cameron-Martin loss achieves a 17% improvement in maximum episode reward (0.95 vs. 0.78 for MSE) and 67.6% reduction in inter-step drifts during inference via the introduced residual magnitude. Similarly, on a 6-station manufacturing line with constant work-in-process (CONWIP) flow control, we achieve 28.4% lower RMSE than classical LSTM baselines; a high starvation-event recall (1.0 in test cycles), and effective bottleneck identification (Precision@1 = 1.0 in test set, 13x signal-to-noise ratio). We then certify the dispatch policies with Hamilton-Jacobi reachability theory which reduces deadlock events by 96% compared to uncontrolled dispatch over 100 simulated runs (351 events prevented).

Temporal difference (TD) learning with linear function approximation is a core method for policy evaluation. Its classical continuous-time description is an ordinary differential equation (ODE), which captures the asymptotic mean dynamics but neglects stochastic fluctuations determining the error floor. We introduce a stochastic differential equation (SDE) approximation for linear TD(0) under Markovian noise. The resulting model distinguishes the contraction dynamics governed by the projected Bellman operator from the influence of Markovian sampling. As a consequence, the model explains the constant-stepsize error floor through the interaction between Markovian long-run covariance and the contraction geometry of the projected Bellman operator.

The shift from video generation to interactive world modeling places new demands on data: beyond captioned videos, world models require temporally aligned video-action-language trajectories grounded in the actions, camera motion, states, and events that drive future scene changes. However, such data is difficult to obtain at scale. Web video datasets offer broad visual coverage but lack executable actions and reliable states; robotic datasets provide action and state supervision but are costly and limited in scene diversity; and existing simulators often lack large-scale human-driven interaction trajectories. In this paper, we introduce EgoCS-400K, a large-scale replay-grounded egocentric Counter-Strike dataset for world models, built from public professional CS and CS2 match demos that preserve human gameplay trajectories and enable parsing, replaying, rendering, and temporal alignment. We extract player states, view directions, movements, keyboard/button inputs, view-angle changes, weapon usage, game events, and round-level context, and render clean first-person videos from the same trajectories. EgoCS-400K contains over 400,000 first-person videos and 10,000 hours of gameplay from more than 1,000 matches and 40,000 rounds, covering 13 maps and 10 player viewpoints per round. It supports a range of interactive visual modeling tasks, including action-conditioned future prediction, state- and event-aware scene rollout, replay-grounded captioning, and agent egocentric action understanding. By connecting visual observations with human actions, camera motion, game states, and events at scale, EgoCS-400K serves as a practical bridge between passive web videos, controllable game simulation, and costly real-world embodied data.

We present a numerical method for the forward solution of nonlinear partial differential equations (PDEs) in which Bellman-Kalaba quasilinearization reduces the nonlinear problem to a sequence of linear subproblems, each discretized by collocation onto a trial space that is linear in its parameters and solved by a single direct linear least-squares QR factorization. The trial space, which we term Linear-in-Learnables (LiL), comprises representations whose trainable parameters enter linearly, including random-feature extreme learning machines, spectral polynomial bases, and trigonometric expansions, each implemented as a physics-informed neural network. The method thus replaces the nonconvex gradient-based training that limits standard PINNs with a convex per-step solve. We establish local Newton-Kantorovich convergence of the outer iteration to a residual-limited neighborhood under an explicit smallness condition, with the limiting accuracy governed by the best-approximation residual of the trial space rather than by an optimization tolerance. The method, denoted LiL-Q, is assessed on seven benchmarks spanning scalar nonlinear PDEs (Bratu, viscous Burgers, Buckley-Leverett), coupled systems (plane-strain elasticity and the incompressible Navier-Stokes equations in two and three spatial dimensions), and steady-state Darcy flow with heterogeneous permeability. Across these problems, LiL-Q converges in single-digit outer iterations in most cases, even at the coarsest basis sizes and independent of the parameter count. When the exact solution lies in the span of the trial space, the method recovers it to machine precision in a single solve. On the Navier-Stokes benchmarks, it matches or exceeds published PINN solvers with up to two orders of magnitude fewer trainable parameters, without gradient-based optimization.

Classifying Cyber Threat Intelligence (CTI) using MITRE Adversarial Tactics, Techniques, and Common Knowledge (ATT&CK) is essential for proactive defense, but historically required extensive human effort. Pre-Large Language Model (LLM) automation sped up this process, but could not resolve the complex language and multi-step attack patterns found in unstructured CTI reports. LLMs addressed previous limitations by using contextual reasoning to understand unstructured text. However, current evaluations rely on simplified, single-technique sentences that ignore the complexity of real-world CTI reports, which often leads to inflated performance results. Consequently, the baseline performance of open-source LLMs on complex unstructured CTI reports remains unevaluated. To address this gap, we constructed a ground-truth dataset of 2,076 human-annotated sentences (1,281 technique-positive, 795 negative) from 83 complex unstructured CTI reports. These sentences were mapped to 114 unique ATT&CK techniques using a six-phase annotation process, achieving \k{appa} = 0.68 inter-annotator agreement. Using this dataset, we evaluated seven open-source LLMs ranging from 8B to 236B parameters across prompt strategy and temperature configurations. The highest-performing LLM achieved a micro-averaged F1 score of 0.22, establishing the empirical baseline for multi-label ATT&CK classification on complex unstructured CTI. Parameter size showed a statistically significant positive correlation with F1 score. Prompt strategy and temperature produced no statistically significant gains across model configurations. These results indicate that current open-source LLMs are insufficient for production-grade ATT&CK classification. The dataset, benchmark, and findings provide a reproducible foundation for future CTI research.

Large language models now produce legal text of at least median quality, yet no existing benchmark can evaluate whether they perform doctrinal legal reasoning, which forms the interpretive core of legal work, rather than the ancillary, paralegal tasks that most current legal-AI evaluations measure. This measurement gap is not only methodological but legal: the EU AI Act makes "appropriate accuracy" a binding requirement for high-risk AI used in the judicial domain, yet that requirement cannot acquire operational content without the very doctrinal-reasoning benchmark the field lacks.

Building personalized cardiac electrophysiology (EP) digital twins requires identifying the appropriate model structure for each patient, not merely fitting parameters. Traditional methods rely on experts to manually prescribe hybrid physics-neural architectures, which requires deep domain expertise and does not transfer across patients. Recent works have applied large language models (LLMs) to generate or act as hybrid models. However, despite their promising generalization capacity, these LLM-based methods lack the structural priors needed for stable cardiac simulations. Hence, we propose LEADS, a framework that formulates cardiac EP domain knowledge as a structured action space and utilizes an LLM agent to discover hybrid models. The agent follows an iterative reasoning-and-action loop to select, combine, and refine hybrid models, whilst gradient descent handles parameter fitting. The proposed LEADS designs every candidate model towards physically grounded, interpretable, and numerically stable, while allowing open-ended architectural discovery. We validate LEADS on synthetic data with three ground-truth reaction models and on real cardiac EP data, demonstrating that it outperforms both human-designed hybrid models and other LLM-based hybrid modeling.

Language models are remarkably capable at medical question answering, in some cases surpassing the accuracy of general physicians. However, answering questions about wearable health data remains challenging and understudied, as these ubiquitous sensors produce continuous, high-dimensional, and longitudinal data, which is non-trivial to align with text-centric distributions in LLM pretraining. The diversity of sensor modalities and user intents cannot be effectively handled by a fixed reasoning workflow or a single pretrained foundation model. To address these challenges, we propose WEQA, a query-adaptive agent framework that unifies LLM reasoning with specialized wearable analytical and modeling tools. An LLM controller is employed to synthesize execution plans and dynamically route each query to the appropriate combination of sensor analysis and pretrained models, and perform grounded response auditing with external knowledge. We also curate a benchmark spanning four open wearable datasets comprising analytic and predictive tasks in three different health domains. Experiments show that our framework is 24% more accurate than LLM and agentic baselines, and a blinded study with 12 medical experts and 8 users shows substantial gains in usefulness and clinical soundness.

A robot's flash endurance is a non-renewable stock: every persisted write spends one of a few thousand program/erase cycles and never refills, yet no fielded robot memory system prices which memories are worth an erase cycle. We treat embodied memory as depreciating capital and price that stock with a single endurance shadow price $η$, which makes cost-minimizing placement across a RAM / on-board NVM / cloud hierarchy a threshold in a wear-augmented per-byte index. The index is cost-optimal whatever the sign of the value-write association $χ$; only when $χ> 0$ does the optimum turn non-monotone, sending a robot's most valuable memories off its flash. The pivot is thus empirical, and we measure $χ$ on real robot logs at a pre-specified gate: its sign is a property of the deployment regime -- positive on recurrent long-horizon manipulation ($\hatχ \approx +1.0 \times 10^{-3}$, replicated at full power), null on a shorter-horizon suite, and negative on non-recurrent teleoperation. Two boundaries scope the result. The endurance budget is dormant on premium 3,000-P/E TLC at datasheet prices and binding on the commodity QLC/eMMC ($\sim$1,000 P/E) that cheaper edge robots run. And where it binds, a learned wear-aware controller only ties price-based routing on task value, because realized value is tier-invariant across RAM, NVM, and cloud: the rent governs device lifetime and cost, not task performance. Whether wear-aware placement improves task value remains open -- $χ$ is measured against a value proxy, and the non-monotone optimum, while proven, is not yet observed in data.

AI agents are moving from advisors to actors, booking travel, planning menus, and running procurement on behalf of users. Existing benchmarks for AI and animal welfare evaluate model text responses to question-answer prompts, leaving open whether the welfare reasoning surfaced in those responses transfers to agentic deployment where the model must take actions with tools. We introduce TAC (Travel Agent Compassion), the first agentic benchmark measuring whether AI agents avoid options involving animal exploitation when acting on behalf of users. TAC presents an AI agent with twelve hand-authored travel booking scenarios across six categories of animal exploitation, augmented to forty-eight samples to control for price, rating, and position confounds. We evaluate seven frontier models from four labs. Every model scores below the chance level of sixty-four percent, with the best performer (Claude Opus 4.7) at fifty-three percent. A single welfare-aware sentence in the system prompt yields gains of forty-seven to sixty-three percentage points in Claude and GPT-5.5, twenty-six points in GPT-5.2, and under twelve points in DeepSeek and Gemini. An auxiliary Inspect Scout audit of 288 base-condition transcripts from the top two performers, using Gemini 2.5 Flash Lite as judge, flags zero transcripts for evaluation awareness, suggesting the below-chance rates do not stem from the models recognising the evaluation. We discuss implications for category-level variation across cultural domains, the limits of text-response welfare benchmarks, and the EU General-Purpose AI Code of Practice systemic risk framework.

Meta-reinforcement learning enables fast adaptation by extracting shared structure from related tasks, but existing end-to-end methods often couple task inference with embodiment-specific control. This coupling can obscure non-parametric task semantics, reduce sample efficiency, and limit cross-agent reuse. We propose a meta-knowledge reutilization framework that learns task-level knowledge on a dynamics-simplified agent and transfers it to heterogeneous agents. The framework uses a Bayesian non-parametric prior to organize latent task modes and a high-level policy to generate task-level magnitude guidance. To bridge reusable task knowledge with different embodiments, we introduce a semantic-magnitude interface and a lightweight temporal adaptor, which convert frozen meta-knowledge into temporally aligned subgoals for embodiment-specific low-level controllers. Experiments on multiple locomotion agents show that our framework reduces final-step tracking error by 94.75% -- 99.79% compared with recent state-of-the-art baselines and achieves comparable deployment performance with about 23.8% of their interaction data.

Recent incidents involving LLMs used for mental-health support reveal a critical evaluation gap: surface-level safety scores do not capture how models behave across realistic, emotionally sensitive interactions over time. Existing benchmarks measure knowledge, safety, or static response quality, but miss whether LLM interactions help users keep reflecting, coping, and making decisions themselves. We formalize this missing dimension as COGNITIVE ATROPHY, a process-level behavioural measure in AI-mediated mental-health support distinct from safety and helpfulness. To measure it, we introduce COGNITIVE ATROPHY BENCH, a clinically grounded benchmark built from 1,576 fully human-generated counseling conversations, 15,680 turns, and 42,230 responses from five LLMs. Three clinical and neuropsychology experts developed a 20-attribute schema spanning user context, response behaviour, and global risk flags; six trained clinical reviewers applied it with span-grounded evidence, producing 5,324 reviewer judgments. We further introduce the User-Input Risk Index (UIRI), the Cognitive Atrophy Risk Index (ARI), and trajectory summaries. Across five LLMs, models show a consistent moderate-to-high level of atrophy-aligned behaviour across single and multi-turn settings. While models generally respond to overt safety cues, they adapt less reliably when users seek solutions or decisions. The dominant recurring patterns are directive advice, problem-solving, recommendation responses, topic shifts, and forms of validation that may reinforce dependence rather than reflection. Our work makes COGNITIVE ATROPHY measurable and provides a foundation for auditing model behaviour in sensitive LLM conversations.

Predicting immune biomarkers associated with the tumor immune microenvironment (TIME) is critical for advancing precision oncology, yet existing approaches are largely limited to single image modalities and suffer from insufficient resolution and incomplete utilization of complementary clinical and biological information. Here we introduce MixTIME, a multimodal foundation model that leverages a mixture-of-experts (MoE) architecture to integrate pathology foundation models trained across distinct modalities: image only (UNIv2), image text (CONCHv1.5), and image transcriptomic (STPath) representations for pixel-level and slide-level prediction of multiplex immunofluorescence (mIF) protein expression from hematoxylin and eosin (HE) whole-slide images. MixTIME employs a learnable router to dynamically weight expert contributions and is trained with a distribution- and tendency-aware loss function. Benchmarked on two datasets of different scales, MixTIME achieves state-of-the-art performance across 17 protein markers as measured by correlation metrics. The predicted mIF profiles substantially enhance downstream tasks, including spatial domain identification, survival prediction, and AI-assisted pathology report generation validated by expert pathologists from multiple institutes across the world. Furthermore, MixTIME enables longitudinal tracking of protein expression dynamics across clinical time points and reveals protein gene interaction patterns linked to drug resistance and immune suppression in tumor microenvironments. Collectively, MixTIME provides a scalable framework for multimodal biomarker discovery and clinical translation in computational pathology.

Modern AI systems increasingly solve a task not with a single model call but with several imperfect agents working together: some propose pieces of a solution, others verify them, and the results are combined. These systems often outperform any single model, yet it is rarely clear why they succeed or when they will fail. We model such a system as message passing on a sparse graph, the structure that underlies low-density parity-check (LDPC) codes, and extend the density-evolution machinery of coding theory to this richer setting. In our model a task is a set of coupled binary subclaims, and an agent architecture is a sparse, role-typed factor graph whose check nodes are noisy Boolean verifier nodes, each computing a local Boolean function of the subclaims it touches. Three distinct failure modes, all modeled as erasures (an agent abstaining, a verifier returning no usable output, and a message lost between two agents), propagate as the agents exchange set-valued messages. The check agents combine these messages by a single logical-forcing rule that specializes to XOR, AND, OR, implication, and Horn constraints. This is more than a relabeling of LDPC theory: the verifier functions are nonlinear and value-asymmetric, and the three failure modes do not reduce to a single effective channel, so they require new threshold, finite-length, and converse results rather than a direct reuse of parity-check density evolution. We prove a density-evolution theorem that predicts the asymptotic fraction of unresolved subclaims on random role-typed architectures, with an extension to deterministic, locally tree-like graph sequences. The XOR case recovers the classical LDPC recursion on the binary erasure channel (BEC); the AND case exposes an asymmetry between positive and negative verifier certificates.

State Space Models (SSMs) such as Mamba-2 offer linear-time inference but their memory footprint limits edge deployment. Prior ternary SSM work (Slender-Mamba) trains from scratch on 150B tokens; we show a pretrained checkpoint suffices, reducing the marginal token budget by 1,000x. Using grouped quantization-aware training (QAT) with knowledge distillation from a frozen FP16 teacher, we compress Mamba-2 1.3B to 3.61x (2,687 to 744 MB) and achieve 48.1% zero-shot accuracy (7-task average) in just 102M tokens (4 GPU-hours, single H100) -- approaching Bi-Mamba's 48.4% (within +/-0.9pp CI). This QAT-from-pretrained setting reveals zero-ratio collapse, a novel instability caused by learnable quantization scales that does not arise in from-scratch training. We further show that post-hoc correction strategies effective for Transformers fail for SSMs due to error accumulation through the recurrence. These results demonstrate that ternary SSMs do not require expensive from-scratch training: QAT from pretrained checkpoints with KD is a data-efficient alternative.

Agentic navigation systems require a base navigation model whose observation strategy can be externally reconfigured at inference time, because instruction following, object search, target tracking, and autonomous driving share the same perception-planning backbone yet demand fundamentally different strategies for consuming the visual stream. We present Qwen-RobotNav, a scalable navigation model built on Qwen-RobotNav that addresses it through a parameterised interface with two complementary dimensions: multiple task modes that select the navigation behaviour, and controllable observation parameters (e.g., token budget, per-camera weights) that govern how visual history is encoded. With training-time randomization over all parameters, Qwen-RobotNav is robust to any inference-time configuration requiring zero architectural modification to the Qwen-RobotNav backbone. We train Qwen-RobotNav on 15.6M samples; co-training with vision-language data prevents the collapse into reactive action-sequence mappers observed in trajectory-only training. The parameterised interface also makes Qwen-RobotNav a natural building block for agentic systems: for long-horizon scenarios, an upper-level planner decomposes goals into sub-tasks and dynamically switches Qwen-RobotNav's task mode and context strategy mid-episode, composing complex behaviours from repeated calls to the same model. Extensive experiments show that Qwen-RobotNav sets new state-of-the-art results across major navigation benchmarks. The model exhibits favourable scaling from 2B to 8B parameters, with joint multi-task training developing a shared spatial-planning substrate that transfers across task families, and demonstrates strong zero-shot generalisation to real-world robots across diverse environments.

We develop a text-to-SQL (structured query language) system based on large language models (LLMs) using in-context learning and apply it to the Automatic Learning for the Rapid Classification of Events (ALeRCE) astronomical database. ALeRCE is a community broker for the Zwicky Transient Facility and the Vera C. Rubin Observatory. The system enables users to query the database in natural language (NL) and generates executable SQL queries. To develop and evaluate the system, we constructed a dataset of 110 NL/SQL pairs. We propose a step-by-step generation framework comprising four modules: schema linking, query classification, prompt decomposition, and self-correction. The performance of thirteen LLMs is evaluated using in-context learning and prompt engineering techniques. Text-to-SQL performance is assessed using the perfect-match (PM) rate for row identifiers (e.g., object identifiers) and column identifiers (i.e., column names). The proposed step-by-step framework consistently outperforms a direct-inference baseline, while the self-correction module consistently reduces execution errors. For Claude Opus 4.6, PM performance on row (column) identifiers is high for simple queries, reaching 0.97 (0.94), and decreases with query complexity to 0.44 (0.72) for medium queries and 0.59 (0.49) for hard queries. Among the thirteen evaluated models, the best-performing LLMs for the text-to-SQL task are Claude Opus 4.6, Gemini 2.5 Pro, Gemini 3 Flash, and GPT-5.2-Codex.

Retrieval-Augmented Generation (RAG) is the prevailing architecture for grounding language model outputs in external evidence, yet its dominant evaluation paradigms and default configurations remain oriented toward factual question-answering. For interpretive disciplines such as historical studies, RAG embeds assumptions that conflict with scholarly practice. We introduce HistoRAG, a framework that translates historiographical principles into concrete architectural interventions. Separated retrieval and generation decouples source discovery from interpretation, temporal windowing enforces balanced source representation across the research period as a methodological requirement of historical inquiry, and LLM-as-judge evaluation makes relevance judgments transparent and contestable. We evaluate these interventions using SPIEGELragged, applied to 102,189 articles from Der Spiegel (1950-1979). Each intervention addresses a measurable deficiency in standard RAG: era-specific vocabulary retrieves zero chunks from the 1950s when using 1970s terminology, evidence of the temporal skew that motivates windowing; vector similarity and LLM-assessed relevance correlate only weakly (Spearman rho = 0.275), motivating post-retrieval evaluation; and keyword-based and semantic retrieval surface largely disjoint source pools, motivating an architecture in which both operate as complementary retrieval layers under a shared LLM evaluation filter. We also introduce the concept of Zwischentexte (intermediate texts that function as interpretive proposals rather than findings) as a framework for responsible integration of LLM-generated text into scholarly practice. The architecture offers a model for how domain-specific epistemological commitments can be translated into RAG design decisions, and may transfer to other interpretive disciplines working with large corpora.

Graphical user interface (GUI) grounding requires vision-language models (VLMs) to identify small target elements in high-resolution screenshots and predict precise screen coordinates. On-policy self-distillation (OPSD) is a promising post-training approach for this coordinate-sensitive task, since it provides dense token-level teacher signals beyond hard coordinate labels. However, naive OPSD is not well suited to GUI grounding: OPSD evaluates the teacher on student-generated prefixes, the quality of coordinate-token teacher signals can degrade when the prefix has already deviated from the target coordinate, leading to unreliable teacher signal. To mitigate this, We propose quality-aware self-distillation for VLM-based GUI grounding, which improves coordinate-token teacher-signal quality through soft correctness-aware gating and teacher-probability scaling. The soft correctness-aware gate checks whether the teacher's current coordinate-token prediction can still be completed into the ground-truth box under the student-generated prefix. If not, the corresponding teacher signal is down-weighted. Teacher-probability scaling then uses the teacher's confidence as a lightweight factor to further calibrate the strength of the gated supervision. A key empirical finding is that neither component alone improves overall performance, whereas combining them consistently improves performance. This suggests that the two mechanisms play complementary roles: correctness-aware gating suppresses unreliable coordinate-token supervision, while teacher-probability scaling calibrates the strength of the remaining signals. Experiments across six GUI grounding benchmarks show that our method consistently improves the base model and outperforms strong baselines.

Advances in Artificial Intelligence (AI) have led AI for Theorem Proving to become a promising means of formally verifying computer systems. Whilst formal verification is traditionally reserved for safety-critical systems due to the required amount of expertise and effort, AI can help to automate a large amount of this workload and make it far more accessible. Blockchain-based systems are becoming increasingly popular and are frequently targeted by malicious actors, often resulting in huge financial losses, highlighting the need to better verify these systems and mitigate vulnerabilities. Arguably the most important component of these systems is the consensus protocol, which allows nodes to agree on decisions in a potentially adversarial environment. In this paper, we improve upon IsabeLLM, the automated theorem proving tool in Isabelle. Namely, we implement a Retrieval-Augmented Generation framework, Error tracing and counterexample generation for improved context supplied to the Large Language Model. Compatibility with the latest version of Isabelle and Sledgehammer is also implemented for improved efficiency. We compare the performance of the two versions of IsabeLLM in their ability to complete the verification of Bitcoin's Proof of Work consensus.

Deformable Linear Objects (DLOs), such as wires and cables, are central to industrial assembly. Unlike rigid objects, whose state is captured by a 6-DoF pose, DLOs have an infinite-dimensional configuration space and deform continuously under contact with grippers, fixtures, and the workspace, making them a demanding benchmark for general dexterous manipulation. Despite their importance, policy development and comparison remain difficult: existing benchmarks are often tied to specific hardware setups, lack modular and customizable task assets, or study generic deformable-object tasks without the fixtures relevant to real-world industrial wire manipulation. Few benchmarks align simulation, real-world data, and shared evaluation protocols. To bridge this gap, we introduce WireCraft, a simulation benchmark for industrial DLO manipulation with configurable difficulty and assets, spanning three task families: connector insertion, clip routing, and channel seating. It supports two complementary DLO physics models, articulated and deformable, and the trajectories come from both simulation and a physical UR5. We benchmark reinforcement learning (RL), imitation learning (IL), and vision-language-action (VLA) policies under shared metrics. Privileged state-based RL solves a representative setting in each task family with over 82\% success, confirming the tasks are well-posed. For connector insertion, however, the transition from reaching the socket to contact-rich alignment remains a key bottleneck for vision RL, IL, and VLA policies. These results indicate that industrial DLO manipulation, though tractable under privileged state, remains an open challenge for current vision-based learning. The benchmark, data, and tools will be open-sourced upon acceptance.

Structured State Space Models (SSMs), including the S4 and S4D architectures, have recently emerged as powerful alternatives to attention-based models for capturing long-range dependencies in sequential data. Despite their strong empirical performance, deploying these models in time- and resource-constrained settings remains challenging due to their computational and memory demands. In this paper, we propose a novel incremental, operator-level pruning approach for S4- and S4D-based models that significantly reduces inference cost while preserving predictive performance. To the best of our knowledge, this is the first work to systematically investigate structured operator pruning for SSMs. Our method progressively prunes model operators by interleaving structured masking with fine-tuning, while jointly monitoring accuracy and inference latency. We implement this approach within a unified training and evaluation framework that enables systematic exploration of efficiency-accuracy trade-offs. Experiments across multiple benchmark datasets show that pruning up to 70% of the model operators preserves the performance of the original models in most cases, while substantially reducing inference latency. These results demonstrate that structured operator pruning is an effective and previously unexplored strategy for improving the efficiency of SSMs and facilitate their deployment in practical, resource-constrained scenarios.

Cross-end-effector grasp generation seeks a unified model that generalizes across objects and across embodiments ranging from parallel grippers to dexterous end effectors. Existing grasp generators are typically designed for a fixed embodiment or encode embodiment identity with a static descriptor, which weakens transfer when topology, actuation coupling, and contact geometry differ substantially. We present EAGG, an embodiment-aligned grasp generator that represents each embodiment with a topology-aware end-effector graph and an embodiment-specific low-dimensional end-effector control space. A frozen end-effector-cognition backbone converts the current articulated state into geometry-aware tokens that act as a reusable morphology prior, and iterative geometry injection refreshes these tokens throughout sampling so that conditioning remains synchronized with the evolving end-effector geometry. On the MultiGripperGrasp benchmark, EAGG reaches 56.17% average success across six training end effectors, remaining within 1.10 percentage points of specialized training while preserving transfer to finetuning and zero-shot end effectors. Iterative geometry injection further reduces the pooled median contact distance from 0.239 cm to 0.189 cm. These results show that cross-end-effector grasp generation is strengthened by aligning embodiment structure inside a shared generator rather than suppressing embodiment differences. Code is available at https://github.com/wanhaoniu/EAGG.

Post-training pipelines that combine supervised fine-tuning (SFT) with reinforcement learning (RL) have emerged as the key recipe for transforming large language models (LLMs) into robust reasoners. We argue that this combined success is driven by compositional generalization, which we formalize through a hierarchical latent selection model. In this framework, reasoning traces are generated by a cascade of discrete latent selection variables corresponding to reusable atomic modules, including both skills (local operations) and routing mechanisms (how intermediate information is selected, reused, and composed). Within this model, we theoretically show that SFT and RL play asymmetric, complementary roles: SFT supplies the raw module materials in compositional traces, and RL decomposes those traces to identify the latent atomic modules and enable compositional generalization. We design controlled experiments to validate this theory. Our results demonstrate that RL can extract atomic modules from compound traces supplied by SFT and recombine them to solve new configurations. Moreover, we find that training on compound traces yields stronger generalization than training on isolated atomic modules. Finally, we investigate the relationship between SFT and RL data and identify an effective protocol in which SFT ensures coverage of all atomic modules through compositional traces, while RL focuses on novel compositions outside the SFT support to drive exploration.

Retrieval-Augmented Generation (RAG) has emerged as a paradigm for enhancing large language models (LLMs) with external knowledge, yet existing graph-based methods face a fundamental limitation: entity-centric and chunk-centric approaches operate on representations anchored to original text without true knowledge fusion. While entity-centric methods connect logically related content and chunk-centric methods preserve context, both retrieve information separately through similarity search, missing emergent understanding from their synthesis. In this paper, we propose HyGRAG, a hierarchical graph RAG framework that transcends source documents by addressing three core challenges: constructing summaries that genuinely integrate contextual and relational information, leveraging these synthesized representations to access emergent knowledge during retrieval, and efficiently updating hierarchical structures for dynamic corpora. Specifically, we design hierarchical index structures over hybrid graphs with both chunk and entity nodes, then iteratively cluster them and generate LLM-based summaries. Then, we design context and relation-aware retrieval that searches across all abstraction levels while expanding through community membership. Moreover, we enable dynamic knowledge update through attachment-based algorithms with only local re-summarization. Experimental results show that HyGRAG improves the average accuracy of multi-hop reasoning tasks by 9.7%, while maintaining reasonable efficiency.

Recent advances in Large Language Models (LLMs) and multi-agent systems have driven the rise of Agentic AI, showing promise for medical reasoning. However, open-ended conversational agents remain prone to two critical failure modes: premature diagnostic handoff and silent clinical hallucinations that may go undetected before reaching the patient. In this work, we propose a multi-agent framework that addresses both issues by replacing ``LLM-as-a-judge'' routing with deterministic orchestration constraints. The framework incorporates two safety mechanisms. First, a neuro-symbolic state-tracking gate enforces completeness of the OLDCARTS clinical protocol (Onset, Location, Duration, Character, Aggravating/Alleviating factors, Radiation, Timing, and Severity) by blocking diagnostic transitions until all required dimensions are collected. Second, an epistemic uncertainty quantification (UQ) gate computes semantic entropy (H) across K=5 independent diagnostic samples to identify and intercept divergent outputs before delivery. We evaluate the system using simulated patient agents powered by the llama-3.1-70b-instruct model on 150 test cases. The full architecture achieves 49.3% diagnostic precision, representing an absolute improvement of 11.3 percentage points over an unconstrained baseline. Additionally, we observe a statistically significant negative correlation (r = -0.181, p < 0.05) between OLDCARTS completeness (σ) and semantic entropy (H), suggesting that structured information gathering is associated with reduced diagnostic uncertainty.

As LLM-driven multi-agent systems (MAS) transition from lab to production, system behavior exhibits nonlinear degradation. We introduce the Intelligence Entropy Principle: probability-driven systems spontaneously drift toward disorder, formalized as S(t) = S0 * exp(alpha*t/Cm), where Cm is a model capability coefficient we propose. Lyapunov analysis yields the stabilization condition lambda > alpha/Cm. We construct the ADE (Agent Delivery Engineering) four-layer framework (L1 Physical Laws through L4 User Adaptation) with 23 core components. Validation spans 100K-scale experiments and 33.6 days of production monitoring. We propose a Five-Layer Disorder Taxonomy unifying failures under structural collapse, and present Elastic Organization as an original MAS morphology. Results: channel fracture reduced from 69-98% to near 0%; system death probability below 0.02%.

As Large Language Model based agents enter autonomous scientific research, their ability to resist pseudoscience becomes increasingly important. Otherwise, such systems may rapidly generate plausible yet misleading studies that contaminate academic literature and erode trust in science. We present PseudoBench, an adversarial benchmark for evaluating whether agentic auto-research systems can identify and resist pseudoscientific narratives. PseudoBench contains 200 curated pseudoscientific claim-evidence pairs across five domains and evaluates agents through an end-to-end research pipeline from experiments to writing. Testing seven state-of-the-art agents, we find that current systems readily produce persuasive reports that align with pseudoscientific premises with near-zero refusal rates and the highest resistance of only 27.4%. Stronger agents risk packaging pseudoscience in more sophisticated scientific language, increasing its apparent credibility. These findings reveal an alarming capacity to fuel pseudoscience, calling for scientific alignment before widespread deployment.

We propose a hybrid grasp synthesis framework that combines a learning-based Energy-Based Model (EBM) with an analytical Iterative Closest Point (ICP) method to generate robust grasps from partially observed point clouds. The learned energy function acts as a prior within a Stein Variational Gradient Descent (SVGD) framework, guiding iterative refinement of grasp configurations. Evaluated on 67 objects with 5,360 grasp attempts, our method achieves an average success rate of 60.9\%, outperforming AnyGrasp (31.1\%) and Grasp Pose Detection (48.4\%) and AS-ICP (56.6\%). These results highlight the strong generalization ability of our approach and demonstrate how combining data-driven learning with geometric optimization addresses the limitations of either strategy in isolation.

LLM agents increasingly rely on external skills -- reusable tool specifications -- but real-world tasks often require composing multiple skills, not just selecting one. We formalize this as the Compositional Skill Routing problem: given a complex user query and a large skill library, decompose the query into atomic sub-tasks, retrieve the appropriate skill for each sub-task, and compose an executable plan. We present SkillWeaver, a decompose-retrieve-compose framework combining an LLM task decomposer, a bi-encoder skill retriever with FAISS indexing, and a dependency-aware DAG planner. To support evaluation, we introduce CompSkillBench, a benchmark of 300 compositional queries over 2,209 real MCP server skills spanning 24 functional categories, sourced from the public MCP ecosystem. Our experiments reveal that task decomposition quality is the primary bottleneck: standard LLM decomposition reaches only 34.2% category recall at the step level. To address this, we propose Iterative Skill-Aware Decomposition (SAD), a retrieval-augmented feedback loop that iteratively aligns decomposition with available skills. SAD improves decomposition accuracy from 51.0% to 67.7% (+32.7%, Wilcoxon p < 10^-6) in a single iteration; DA-conditioned analysis confirms that correct granularity is the prerequisite for effective retrieval (CatR@1 rises from 34% to 41% when DA=1). SkillWeaver reduces context window consumption by over 99%, and transfer experiments confirm generalization (+35.6% relative DA gain even when target categories are absent from the retrieval pool).

Decision-making with deep learning-based time series forecasting requires not only accurate predictions but also actionable insights. However, current architectures do not inherently provide such information. Specifically, guidance is needed on how current conditions must be modified to shift from a predicted outcome to a desired future scenario. Counterfactual explanations provide a natural framework for this task, as they represent minimal input changes that alter the model's prediction, indicating when and how intervention is required. Existing approaches rely on instance-wise optimization, leading to inconsistency across instances, high computational costs, and limited applicability in real-time settings. To address these limitations, we reformulate counterfactual generation for time series forecasting as the problem of learning a globally consistent intervention strategy, allowing counterfactuals to be generated through a single shared function. We propose Counterfactual Time Series Explanations (ConTex), a model-agnostic, decomposed architecture comprising a temporal context encoder and a conditional encoder, followed by two heads that capture interventions in terms of temporal relevance and modification strength. This structure overcomes the instability and inconsistency of instance-based approaches by producing targeted, interpretable interventions across time and feature dimensions in a single forward pass, making it suitable for real-time applications. Across multiple forecasting architectures and benchmark datasets, ConTex achieves state-of-the-art validity while generating sparse counterfactuals that minimize the number of necessary interventions. Additionally, our approach reduces computational cost by at least 12-36x compared to instance-wise generation and supports real-time inference at approximately 0.007 seconds.

Vision-language-action models (VLAs) combine vision-language backbones with expressive generative action heads trained via flow matching on large-scale robotic datasets. Despite their strong empirical performance in robotic manipulation, VLAs lack mechanisms to quantify confidence in their predictions and to detect when their actions may be unreliable. This presents a critical limitation for real-world deployment in non-stationary environments, where models inevitably encounter scenarios outside their pretraining distribution and may fail without warning. To address this, we derive an efficient method for quantifying epistemic uncertainty in flow-matching models by leveraging velocity-field disagreement (VFD) across a small ensemble. We successfully use this uncertainty estimate for failure detection during deployment and active fine-tuning of flow-based VLAs. To this end, we propose SAVE, a framework for uncertainty-guided active multitask fine-tuning that reduces the number of costly expert demonstrations required to adapt VLAs to new tasks. Through extensive experiments on the LIBERO benchmark, we demonstrate that VFD yields better-calibrated uncertainty estimates predictive of downstream performance, that VFD achieves strong performance in detecting failures, and that uncertainty-guided data acquisition with SAVE requires at least 22% fewer samples than baselines. In summary, our work shows that quantifying epistemic uncertainty in flow-based VLAs improves both failure awareness and adaptation. Project website: tum-lsy.github.io/uq_vla/.

Tool-using LLM agents increasingly use the Model Context Protocol (MCP) to answer from heterogeneous evidence sources, including search, APIs, databases, clinical records, and formulary tools. Standard factuality metrics usually test whether an answer is supported by pooled evidence, missing a provenance-sensitive failure mode: a claim may be supported somewhere while being attributed to the wrong source. We call this cross-source conflation. We introduce ProvenanceGuard, a source-aware verifier for MCP-grounded answers. It consumes captured MCP traces with stable tool IDs, source IDs, and raw outputs; decomposes answers into atomic claims; routes claims to source-specific evidence; checks support with NLI and a token-alignment proxy; compares stated attribution with the routed source; and returns per-claim verdicts plus an answer-level allow/block decision. Blocked answers can be repaired with retrieval-augmented answer revision and re-verified. We evaluate on 281 medical-domain MCP-agent traces. A 266-trace adjudicated subset yields 2,325 LLM-assisted claim labels split by trace; 361 held-out labels are human-verified. On the 40-trace held-out split, ProvenanceGuard achieves block F1 0.802 and source accuracy 0.858 over 260 source-eligible claims, outperforming source-blind baselines that do not emit claim-to-source IDs. On a harder multi-source benchmark it reaches block F1 0.846, while source-plus-relation accuracy drops to 0.229, showing that exact source ownership remains difficult with semantically close sources. Repair-and-reverify resolves all blocked answers in the full trace set, often via conservative fallback. In 50 controlled clinical conflation probes, ProvenanceGuard detects all injected attribution swaps with no retained wrong attribution. These results show that source attribution is an independent axis for factuality verification in MCP-based agents.

Looped Transformers scale latent computation by repeatedly applying shared blocks, but sequential looping increases latency and KV-cache memory with the loop count. Parallel loop Transformers (PLT) alleviate this cost through cross-loop position offsets (CLP) and shared-KV gated sliding-window attention, making loop count a practical design choice. We therefore study PLT loop-count selection through a gain--cost view: an extra loop may refine representations, but CLP also introduces a positional mismatch at each loop boundary. We instantiate this study by training LoopCoder-v2, a family of 7B PLT coders with different loop counts, from scratch on 18T tokens, followed by matched instruction tuning and evaluation. Empirically, the two-loop variant delivers broad gains over the non-looped baseline across code generation, code reasoning, agentic software engineering, and tool-use benchmarks, improving SWE-bench Verified from 43.0 to 64.4 points and Multi-SWE from 14.0 to 31.0 points. In contrast, variants with three or more loops regress, revealing a strongly non-monotonic loop-count effect. Our diagnostics show that loop 2 provides the main productive refinement, while later loops yield diminishing, oscillatory updates and reduced representational diversity. Because the CLP-induced mismatch remains roughly fixed as refinement gains shrink, the offset cost increasingly dominates. This gain--cost trade-off explains PLT's saturation at two loops and provides diagnostics for loop-count selection.

Masked diffusion models (MDMs) have recently emerged as a promising paradigm for sequence generation. Scaling MDMs is conventionally achieved by increasing the parameter count or the number of denoising steps. We introduce Recursive Masked Diffusion Models (R-MDMs), which add recursive depth as a third scaling axis by repeatedly applying the same denoising transformer within each diffusion step. Recursion enables iterative refinement of the output through parameter reuse, increasing effective model depth without increasing parameter count. Across structured generation tasks, including Sudoku and Countdown, we show that R-MDMs achieve substantially improved parameter efficiency: a model with $L$ recursive iterations often matches the performance of non-recursive baselines with roughly $L\times$ more parameters. Moreover, recursive refinement can partially substitute for additional denoising steps, allowing recursive models to reach the same generation quality with fewer forward passes at inference time. These results suggest that recursive depth is a practically useful scaling mechanism for MDMs, improving both parameter efficiency and the allocation of test-time compute.

AI systems deployed in legal workflows hallucinate at rates that aggregate metrics report at ~52%, but this average conceals where errors concentrate and in which direction they run, leaving compliance officers without an actionable signal for trustworthy deployment. We present LegalHalluLens, an auditing framework with three components: typed hallucination profiles across four legally-motivated claim categories (numeric, temporal, obligation/entitlement, factual) over CUAD (Hendrycks et al., 2021); a Risk Direction Index (RDI) that reduces omission-versus-invention bias to a single deployment-comparable scalar; and a typed debate pipeline calibrated to both magnitudes and directions. Across 510 contracts and 249,252 clause-level instances we measure a within-model gap of approximately 38-40 pp between obligation/numeric and temporal claims that aggregate reporting hides, and show that two systems with matched 52% rates can carry opposite RDIs. The debate pipeline reduces fabricated detections by 45% with per-category gains tracking the diagnosis, matching commercial APIs with a substantially smaller backbone (4B active parameters). Typed profiles and RDI surface failure modes that aggregate metrics hide; we further show these diagnostics serve as calibration inputs for multi-agent debate pipelines, where Skeptic challenges and asymmetric gates targeted at measured failure modes outperform generically-tuned debate. The framework supports direction-aware procurement, accountability, and agent design for legal AI deployed in the wild.

Large language models (LLMs) are increasingly deployed as autonomous agents that make consumption decisions on behalf of users. This shift raises fundamental questions for consumer theory, which has traditionally modeled humans as the primary decision-makers. In this paper, we introduce LLM Consumer Behavior Theory, a new field of study concerned with analyzing consumer behavior in agentic markets. Drawing on classical and behavioral economics alongside recent advances in Natural Language Processing, we formalize how human preferences are reflected and acted upon by LLM-based agents, and how agent-level decisions aggregate into market demand. We unify previously fragmented literature on LLM decision-making, human behavior simulation, and preference elicitation under a common economic lens, highlighting where assumptions, such as rationality and heterogeneity, may fail in agentic markets. Rather than providing empirical validation, this paper outlines the scope of LLM consumer behavior and identifies open research questions related to alignment, preference representation, and market dynamics.

Knowledge graph (KG) foundation models (KGFMs) are zero-shot generalizers: trained once, they can predict links on unseen graphs without retraining. However, understanding when and how they can robustly generalize across KGs is still an open question. In this paper, we shed some light on their generalization mechanisms highlighting how their performance on unseen KGs is not uniform when it comes to partially seen links, which we call half-links. In fact, we show that to predict a test triple $(h,r,t)$ it might suffice in practice to have observed the half-link $(h,r)$ or $(r,t)$ in the inference graph. This yields a taxonomy of four scenarios when combinations of these half-links are observed or not. In a rigorous stratified analysis over these scenarios, we reveal that SoTA KGFMs use seen half links for predictions, while unseen half-links pose different challenges. As such, our finer-grained taxonomy can be a diagnostic protocol for robust KGFM generalization and highlights where novel KGFMs can improve.

Optical networks need intent-driven, closed-loop agentic management, a key enabler for higher autonomy levels. We present the first T-API-compliant reasoning and act (ReAct) loop. We show that domain-specific composite tools achieve 90% oracle-validated correctness with threefold token savings compared to generic tools.

MDLMs generate text by denoising a preallocated masked response canvas, making response-length modeling central to instruction tuning. Existing MDLMs often inherit the autoregressive convention of using repeated \texttt{[EOS]} tokens for padding during instruction tuning, giving \texttt{[EOS]} a dual role as both a semantic terminator and a padding token. We show that this dual role is a root cause of \texttt{[EOS]} overflow under large-block decoding. To decouple these roles, we propose VoidPadding, which introduces \texttt{[VOID]} for padding and reserves \texttt{[EOS]} for termination. During inference, the learned \texttt{[EOS]} signal enables early stopping, while the learned \texttt{[VOID]} signal guides adaptive response canvas expansion. On Dream-7B-Instruct, VoidPadding improves the block-size-averaged four-task mean across mathematical reasoning and code generation benchmarks by \(+17.84\) points over the original model and \(+6.95\) points over RainbowPadding, while reducing decoding NFE by 55.7\% on average. Code is available at https://github.com/Haru-LCY/VoidPadding.

Existing low-light image enhancement methods often face a bottleneck between the representation capacity of illumination-field modeling and computational complexity. To address this issue, this paper proposes an Adaptive Illumination Gaussian Splatting Network (AIGS-Net), an ultra-lightweight architecture for fast low-light enhancement. Unlike conventional static priors, AIGS-Net constructs an input-adaptive 2D Gaussian Splatting illumination field. The opacity of Gaussian basis functions is dynamically modulated by relative luminance statistics of the input image, and spatially varying illumination compensation is rendered through ordered alpha compositing. To guide adaptive illumination compensation efficiently, a zero-parameter nonlinear multiscale contextual encoding module is introduced to extract low-frequency structures and local contrast cues without additional convolutional weights. To suppress noise amplification and sensor-induced color bias, AIGS-Net integrates noise-mask estimation, locked single-channel Gamma mapping, cross-channel consistency regularization, and target color-alignment constraints. Experiments on LOL and LSRW benchmarks show that AIGS-Net improves detail recovery and color fidelity while requiring only approximately 40 learnable parameters, achieving an effective trade-off between enhancement quality and extreme inference efficiency.

Diffusion-based visuomotor policies deployed with asynchronous inference often exhibit inter-chunk discontinuities and lack explicit mechanisms for obstacle-aware execution, leading to jerky motions and collisions that hinder reliable manipulation in real-world scenes. To address these issues, we propose LAGO Policy, a unified asynchronous action-generation framework that integrates trajectory optimization with diffusion policy for smooth and safe execution. LAGO Policy improves inter-chunk consistency via latency-aware classifier-free guidance conditioning on future actions. It further enables goal-directed collision-free trajectory planning by predicting a task-relevant interaction goal from demonstrations. Finally, spatial-temporal trajectory optimization refines the actions to be executed for low-jerk and feasible motion. Extensive real-world experiments demonstrate that LAGO Policy achieves smooth collision-free execution with high task success across challenging manipulation tasks. Project Website: https://lago-policy.github.io/

Existing RL post-training methods for text-to-image generation usually convert the final-image reward into a single scalar advantage and apply it with the same strength to the entire generative trajectory. However, text-to-image generation naturally has temporal and spatial structure: different denoising steps are responsible for different generation stages, and the content that truly determines text alignment often appears only in part of the image. This granularity mismatch makes it difficult for policy updates to focus on the generative components that actually affect the reward. To address this issue, we propose \textbf{SpatioTemporal Adaptive Reward (STAR) Allocation} for RL post-training of text-to-image diffusion and flow models. STAR uses text-image attention inside the generative model and starts from the core content that the user truly cares about in the prompt. It constructs spatial allocation maps that dynamically vary across denoising steps and rollouts, and allocates the same group-relative advantage to more relevant latent regions with almost no additional computational overhead. STAR then applies stronger policy updates to these regions through a spatially resolved policy objective. We use Stable Diffusion 3.5 Medium as the base model and evaluate on three tasks: GenEval, OCR text rendering, and PickScore. Experimental results show that STAR improves compositional semantic alignment, text rendering, and preference optimization without changing the external reward source, achieving $\mathbf{0.9759}$, $\mathbf{0.9757}$, and $\mathbf{23.60}$ on GenEval, OCR, and PickScore, respectively.

Trajectory similarity is a fundamental task in analyzing mobility patterns, essential for applications such as route pattern extraction, mobility prediction, and anomaly detection. Traditional distance-based measures for computing similarity incur high computational cost, driving the adoption of lightweight learning-based approaches. Supervised methods rely on extensive labels derived from traditional distance measures and often reproduce these metrics, which limits generalization. While self-supervised learning addresses this issue through contrastive learning, it lacks a unified framework, making it difficult to compare deep learning (DL) models for consistent trajectory representation. Accordingly, this paper presents MoCo-AIS, a unified framework for learning vessel trajectory embeddings based on the Momentum Contrast (MoCo) paradigm, which formulates similarity learning through positive and negative trajectory pairs. Within this framework, we evaluate a diverse set of leading DL models on large-scale, real-world vessel-tracking AIS datasets that capture diverse navigation behaviors and operating conditions. Results demonstrate that our framework significantly improves similarity learning over existing baselines, while providing a benchmarking platform for evaluating trajectory representation models.

Self-supervised DINO models provide strong transferable visual representations, yet applying them directly to image segmentation remains challenging. Existing approaches commonly rely on heavy decoders with complex upsampling, introducing substantial parameter and computational overhead. We observe that introducing scale into DINO features is far more critical than increasing decoder capacity. In this work, we present SegDINO, an efficient segmentation framework that integrates a DINOv3 backbone with lightweight scale modeling. SegDINO introduces Token Pyramid Adaptation (TPA) to reorganize intermediate DINO features into a pseudo multi-scale hierarchy, and Scale-Aware Decoding (SAD) for efficient intra-scale refinement and top-down multi-scale propagation. We further curate PanCT, a new CT dataset containing 284 patients with expert-annotated pancreatic tumors, to assess SegDINO's ability to handle difficult small-lesion cases. Extensive experiments on PanCT and three public benchmarks demonstrate that SegDINO achieves state-of-the-art results with high efficiency. The code is available at https://github.com/script-Yang/segdino_v2.

Speech foundation models, pre-trained on large corpora of unlabelled speech data, produce general-purpose representations which are useful across tasks. However, these representations encode information about salient speech variables in a distributed manner, while downstream speech tasks rely on only some of this variability. In this work, we propose a post-training refinement approach using interventional contrastive learning. By leveraging an interventional dataset and multi-part contrastive loss, we learn a transformation from the entangled representation space of speech foundation models into separate content and speaker subspaces. We evaluate the learnt representations on speaker verification and keyword spotting tasks, showing improved out-of-domain speaker verification performance and evidence that speaker and content information are separated across the learned subspaces.

Reasoning about what agents can achieve through strategic interaction is a core challenge in Multi-Agent Systems (MAS). Logics for strategic ability, such as ATL, provide rigorous methods, but their adoption is often hindered by the computational cost of strategy synthesis. We introduce a neuro-symbolic framework that integrates large language models (LLMs) into the model-checking pipeline for MAS. The LLM acts as a strategy-generation oracle, proposing candidate strategies that are then formally validated by a standard MAS model checker. This generate-and-certify architecture uses LLM guidance to navigate large combinatorial strategy spaces while preserving formal soundness: generated strategies are accepted only when certified by the verifier. We instantiate the framework for bounded strategic reasoning in NatATL and introduce the first NatATL strategy-synthesis dataset, consisting of 4211 instances. Experiments with an open-weight Qwen3-32B model show that our certified pipeline achieves 92\% accuracy on strategy-synthesis outcomes.

Positional encoding is a fundamental component of Transformer architectures, as it injects information about the spatial or sequential arrangement of inputs. Among recent alternatives to standard absolute and sinusoidal encodings, similarity-based positional encoding (simPE) has emerged as a flexible framework for representing positional structure through pairwise relations. simPE was originally designed for medical imaging applications, where geometric robustness is especially relevant: small rotations naturally arise during image acquisition, induced by imaging instruments, patient positioning, or slight acquisition misalignments. Despite its empirical promise, the theoretical behavior of simPE under geometric perturbations has not been fully characterized. In this paper, we study the robustness of simPE with respect to rotations, combining formal theoretical analysis with experimental validation. We first show that simPE is generally not rotation-invariant. We then prove that, under mild Lipschitz assumptions on the elementary components, simPE is stable under rotational perturbations and derive explicit perturbation bounds in Frobenius norm. We validate these findings experimentally on four controlled datasets--a synthetic Arrow dataset, a synthetic Shapes dataset (four geometric shape categories), a synthetic Digits dataset, and a benchmark image classification dataset (FashionMNIST)--in which training and validation images are kept in a fixed canonical orientation while test images are subjected to increasing rotation angles. Across all datasets, simPE consistently outperforms standard learned positional encoding in terms of accuracy, F1 score, precision, and recall under rotation, particularly in the small-to-moderate angle regime, corroborating the theoretical stability guarantees.

When vision contradicts text, multimodal large language models (MLLMs) consistently favor text, even when images provide clear evidence otherwise. This bias poses risks for applications requiring visual grounding, yet its cause remains unclear. In this paper, we uncover a surprising finding: models often get it right initially, forming correct vision-based predictions in their intermediate layers, before changing their minds and favoring text in the final output. We call this "late-layer textual override". The visual information is encoded, it simply does not survive to the output. More intriguingly, we find that how predictions change reveals whether they're correct: 85% of failures shift toward text, while 89% of successes shift toward vision. This directional signature enables a simple but powerful intervention: when we detect a confident visual prediction being suppressed, we restore it. We propose CALRD (Conflict-Aware Layer Reference Decoding), a training-free method that recovers overridden predictions at inference time. Experiments across five MLLMs of varying architectures demonstrate up to 9.4% absolute improvements on conflict benchmarks while largely preserving standard performance, without training or external knowledge. It recovers what the model already knew but failed to preserve.

Visual information helps resolve ambiguity in coreference resolution, leading to notable performance gains. However, existing Multi-modal Coreference Resolution (MCR) methods require training with (partially) annotated data from the target dataset before they can be applied, preventing their direct usability and raising concerns about generalization. While Vision-Language Large Models (VLLMs) with billions of parameters offer promising zero-shot capabilities, they remain largely inaccessible. Their massive size limits deployability, and many are only accessible through paid APIs. In this paper, we propose a plug-and-adapt method that strategically adapts a carefully pre-trained \emph{alignment model} for immediate use in MCR tasks, designed to eliminate the need for training on scarce benchmark datasets or relying on resource-intensive VLLMs. Specifically, we first pre-train a fine-grained alignment model between textual and visual contextual information using vision-language alignment datasets. We then repurpose the alignment model to MCR through similarity aggregation by fusing visual and categorical cues with evidence theory, thereby enhancing effectiveness. Experiments on the Coreference Image Narratives (CIN) benchmark dataset demonstrate the effectiveness of our method, achieving a 5.31\% and 2.12\% improvement in CoNLL F1 over SOTA dedicated methods and popular VLLMs, respectively. We further evaluate our method on a masked CIN dataset for robustness testing and on a specially constructed VCR-MCR dataset for generalization assessment, with results confirming both capabilities.

Most Vision-Language-Action (VLA) models map observations directly to actions without explicit reasoning, limiting their capacity for reasoning-intensive long-horizon tasks. To address this, existing approaches adopt Chain-of-Thought (CoT) reasoning to enable subgoal decomposition and spatial anticipation. However, those methods lack a unified architecture for effective cross-modal reasoning and fail to explicitly include inverse reasoning ability based on the target state. We argue that manipulation planning naturally decomposes into prediction, anticipating the next visual state, and inverse dynamics, inferring the actions to reach it. Bridging both requires a unified autoregressive architecture that interleaves textual and visual reasoning in a single generation process. We propose \textbf{ThinkingVLA}, a generative model that realizes this decomposition within a unified Mixture-of-Transformers architecture. ThinkingVLA consists of a forward CoT that identifies the immediate subgoal and guides the visual forecasting; the predicted image then serves as the target state, grounding an inverse CoT that reasons about spatial relationships and action intent based on the predicted image; and the final action is generated conditioned on this full reasoning context. Extensive experiments on simulation and real-world benchmarks demonstrate that ThinkingVLA consistently outperforms state-of-the-art baselines, with particularly large gains on long-horizon manipulation tasks.

AI evaluations are shifting toward harder tasks that benefit from longer trajectories involving tool use and iterative problem solving. As a result, performance is increasingly sensitive to the amount and allocation of compute available at test time ("inference compute"). Yet many evaluations still report performance at a single restrictive budget, meaning that low scores may reflect the evaluation setup rather than the model's underlying capability. To test this, we evaluate up to 12 frontier language models on seven challenging benchmarks spanning software engineering, mathematics, medicine, and cybersecurity. We use a controlled setup combining three simple inference-scaling interventions: larger token budgets, context compaction, and repeated submission attempts, guided either by the model itself or by minimal correctness feedback. We find three main results. First, larger token budgets substantially improve performance on benchmarks across multiple domains, including cybersecurity, FrontierMath, Humanity's Last Exam, and TerminalBench. Second, fixed-budget evaluations can increasingly understate frontier capability as models advance. Newer models reach higher performance at large budgets, where they unlock harder tasks and solve them more reliably. Third, benchmarks differ in which inference-scaling methods help most: repeated submission broadly improves performance, but the value of larger token budgets, external feedback, and parallel attempts varies by benchmark. Overall, our results show that benchmark scores are protocol-dependent. We therefore argue that evaluations should report capability as a function of inference-time compute, specify protocol choices explicitly, and compare model generations over a large shared compute range at matched budgets, especially in safety- or policy-relevant settings.

Computer-using agents drive real software through the screen -- clicking and typing -- but they solve every task from scratch: asked to repeat a task, an agent re-reads the screen, re-reasons every tap, and pays the full cost again. We present PreAct, which lets such an agent get faster on tasks it has done before. The first time it succeeds, PreAct compiles the run into a small state-machine program-states that check the screen, transitions that act-and on later runs replays it directly instead of invoking the agent 8.5-13x faster, with no per-step language-model calls. Replay is not blind: at each step PreAct checks that the screen matches what the program expects before acting, and hands control back to the agent the moment something is off. PreAct applies the same discipline when deciding what to keep: a freshly compiled program enters the store only if, re-run from a clean state, an independent evaluator confirms it solved the task-catching programs that replay to their last step yet leave the task undone. Across a mobile, a desktop, and a web benchmark, this store-time check separates repeated runs that improve from ones that degrade as faulty programs accumulate, worth 1.75-2.6 tasks per benchmark, the same direction on all three; a fallback that explores afresh when no program fits brings PreAct level with a strong record-and-replay baseline. We also report what did not matter: prompt wording, runtime guardrails, and whether a language model or a plain embedding retriever selects which program to reuse.

Kolmogorov-Arnold Networks (KANs) have recently emerged as a promising alternative to traditional multilayer perceptrons by replacing linear weights with learnable univariate functions. Despite their theoretical advantages in interpretability and expressiveness, practical research of KANs remains difficult due to high computational costs and inconsistent feature support across existing frameworks. This paper introduces KANLib, a modular, extensible, and computationally efficient framework for developing and evaluating KAN architectures. KANLib unifies core concepts from existing implementations, including PyKAN, EfficientKAN, and FastKAN, within a consistent software architecture that emphasizes flexibility, feature parity, and high performance. The framework supports two basis function types, adaptive grid rescaling, grid extension, and fine-grained architectural customization while maintaining compatibility with standard PyTorch workflows. Experimental evaluation on the California Housing benchmark demonstrates that KANLib reproduces the predictive behavior of established reference KAN implementations while achieving competitive computational efficiency. Furthermore, the framework enables the exploration of architectural variations beyond standard KAN formulations with only minor impacts on predictive performance. Overall, KANLib provides a robust foundation for future research on scalable and extensible KAN architectures.

Current Vision-Language-Action (VLA) models face a trade-off between efficient action generation and explicit deliberation. Directly decoding actions from vision-language backbone representations enables low-latency control, whereas explicit reasoning through textual chains, pixel-level subgoals, or action search can improve planning but incurs substantial latency and computational cost. We propose PearlVLA, a VLA framework that moves deliberation into the latent space of a vision-language model (VLM). PearlVLA separates VLM meta-query representations into a fixed visual grounding branch and an iterative latent plan branch. At each refinement round, a plan-conditioned world query probes a lightweight frozen latent world model for an action-free future observation latent, which is fed back to guide plan refinement. A future-guided RefineNet then applies scheduled residual updates to progressively refine a coarse semantic draft into a fine-grained latent action plan. The refined plan after K rounds is then decoded in parallel into an action chunk for low-latency execution. We further introduce Causal Refinement-Grouped Process-Reward RL to optimize the latent refinement process with rewards from longer-horizon imagined futures induced by latent plan edits. Empirical evaluations on the LIBERO benchmark demonstrate that PearlVLA achieves state-of-the-art performance among existing methods.

Big-Data-as-a-Service (BDaaS) platforms require re liable automation across data ingestion, cleaning, feature engi neering, model development, deployment, and post-deployment monitoring. However, existing LLM-based data science agents and AutoML systems mainly focus on isolated workflow stages, leaving limited support for lifecycle-level orchestration, artifact governance, human oversight, and drift-aware adaptation. This paper proposes a trustworthy self-composable BDaaS frame work based on LLM-orchestrated multi-agent collaboration. The proposed architecture decomposes the BDaaS lifecycle into specialized agents for data ingestion, data cleaning, feature engineering, AutoML training, model evaluation, MLOps de ployment, monitoring, and drift detection. A central LLM or chestration layer coordinates agent execution, validates interme diate outputs, manages workflow context, and enables dynamic workflow composition. The framework also incorporates shared artifact governance, reproducibility support, human-in-the-loop checkpoints, and drift-aware feedback loops. A prototype-based evaluation is conducted using controlled tabular benchmark datasets with missing values, categorical variables, outliers, class imbalance, and simulated covariate drift. Compared with manual ML, AutoML-only, and single-agent LLM baselines, the pro posed multi-agent BDaaS pipeline achieves competitive predictive performance while improving lifecycle-level reliability, including workflow completion, artifact traceability, deployment readiness, reproducibility, and drift recovery. The results suggest that LLM-orchestrated multi-agent systems can extend conventional AutoML toward trustworthy, adaptive, and production-oriented BDaaS lifecycle automation.

Dense retrieval has become the dominant paradigm in information retrieval, in which each document is scored against a query by the inner product of their vector embeddings, and the top-$k$ documents by score are retrieved for this query. However, since each document's score depends solely on the embedding of the query and itself, the retrieval process is oblivious to the content of the entire corpus. Therefore, dense retrieval cannot avoid selecting semantically similar documents from the corpus, which may result in a non-diverse, redundant set of retrieved documents. To this end, we approach retrieval as a joint decoding problem, in which documents are selected as a set with regard to the context of the rest of the corpus. To achieve this, we propose Non-Negative elastic Net (NNN) decoding, which selects documents whose embeddings jointly reconstruct the query embedding as a sparse non-negative linear combination. Our main theoretical result establishes a strict separation between dense retrieval and NNN decoding. For any corpus, every query correctly handled by dense retrieval is also handled by NNN decoding, while on corpora containing correlated documents, NNN decoding additionally handles queries that dense retrieval cannot. Experimental results indicate that applying NNN decoding to frozen embeddings trained for inner-product scoring yields consistent improvements across several benchmarks. Moreover, we introduce an end-to-end training procedure which optimizes the embeddings for NNN decoding, producing significant performance gains surpassing in all metrics and benchmarks compared to dense retrieval. Our work establishes a new paradigm for leveraging dense embeddings in information retrieval, beyond the standard practice of inner-product scoring.

Recent World-Action (WA) models demonstrate strong generalization ability and data efficiency, but they typically rely on expert trajectories for training. This reliance limits their ability to acquire fine-grained manipulation skills beyond the demonstration distribution and prevents them from continuously improving through real-world interaction. To address these limitations, we propose WAM-RL, a reinforcement learning framework that enables joint optimization of the world model and the action model through online interaction with the environment. By allowing the two components to co-evolve, our approach enhances fine-grained control and adaptability. Specifically, a WA model consists of a world model and an actor. We design a tailored reinforcement learning method with hierarchical optimization to coordinate their improvement. On the methodological side, we systematically investigate the effects of applying reinforcement learning to the action model, as well as online training of the world model within an RL setting. Our experiments reveal a key insight: optimizing only the actor yields improvements on short-horizon tasks, but fails to provide significant gains on long-horizon tasks. In contrast, jointly optimizing both the world model and the actor is critical for achieving strong performance in long-horizon settings. Our work is the first to introduce reinforcement learning into the World-Action paradigm, and provides insights into how online optimization of both the action head and the world model impacts overall performance.

Large language models perform increasingly well on standardized logical reasoning benchmarks, but whether this ability remains robust beyond English is unclear. We introduce ChLogic, an English--Chinese aligned benchmark that tests whether models preserve logical reasoning performance when the same latent logical structure is expressed in English and diverse Chinese surface realizations. Built from formal logical templates, the benchmark contains three data sets: (i) the General aligned set, derived from 60 General Propositions across nine template families; (ii) the Difficult aligned set, derived from 40 Difficult Problems; and (iii) the Chinese-only set, covering 15 language-specific phenomenon types. Each aligned item pairs one English reference expression with five Chinese realizations. Experiments on Qwen3, Ministral, and GLM models reveal a persistent English--Chinese performance gap. Back-translation from standard Chinese into English often improves performance on the General aligned set, but produces mixed effects on the Difficult aligned set, where Qwen3-32B and GLM-5.1 perform worse after translation. These results indicate that Chinese surface realization, translation artifacts, and model-specific behavior jointly affect multilingual logical reasoning. Overall, ChLogic provides a useful stress test for the robustness of multilingual reasoning.

Language models increasingly serve as advisory systems in maintenance operations. To prevent hallucination, recent systems ground these models in procedural documentation to constrain them to approved steps. In practice, however, operator queries frequently stray from this path, requiring models to recognise out-of-scope inputs mid-conversation, a dynamic that current benchmarks rarely prioritise. We introduce DiagFlowBench, a dataset of 50 industrial diagnostic flowcharts from a consumer manufacturer converted into 1,676 multi-turn conversations that contrast compliant with out-of-scope utterances. Evaluating a panel of ten commercial and open-weight models reveals high variability in abstention rates, with models commonly selecting a real but contextually inadequate step rather than fabricating facts. The inherent plausibility and authority of this mapped but wrong advice exposes a challenging vulnerability for grounding systems.

Long-term human path forecasting in crowds is critical for autonomous moving platforms (like autonomous driving cars and social robots) to avoid collision and make high-quality planning. Although the current research take into account social interactions for prediction, they don't reveal the exact kinds of social interactions happened among people and how the social interactions affect the decision-making process of pedestrians, which further limits its robustness. Social interactions in pedestrian walking are intuitively massive and hard to label and quantify. In this paper, we explore creatively to quantify and interpret how pedestrians interact with others by proposing Learn to Cluster. Our clustering social interactions is probabilistic latent variable generative, learning directly from sequential trajectory observations, scalable to arbitrary number of pedestrians. Learn to cluster is label-free and can be naturally integrated into the training process of the prediction model. The latent variables will then serve as 'labels' to categorize social interactions. Extensive experiments over several trajectory prediction benchmarks demonstrate that our method is able to learn the patterns of social interactions and effectively integrate the patterns to pedestrian trajectory prediction.

Long-form chain-of-thought reasoning can improve LLM performance on complex tasks, but models often continue generating unnecessary reasoning after a correct answer has emerged. We refer to this behavior as overthinking. We study this phenomenon from the perspective of GRPO-style reinforcement learning (RL) post-training, framing it as a training-time credit-assignment problem rather than merely a decoding-time stopping problem. In rollouts sampled at the onset of GRPO training, we observe that successful trajectories can exhibit a slightly higher degree of overthinking than unsuccessful trajectories for the same prompts. This early imbalance provides a starting point for an undesirable feedback loop: because GRPO assigns sequence-level credit, it cannot distinguish the solution-reaching prefix from the unnecessary continuation that lengthens a successful trajectory. Both receive positive update signal, allowing the initial imbalance to grow into more severe overthinking during training. To address this issue, we introduce Dynamic Rollout Editing (DRE), a training-time intervention for successful trajectories that continue thinking after answer emergence. DRE preserves the accepted verified prefix, edits the remaining thinking, and prefers the edited trajectory within the same RL group, weakening the preference signal for unnecessary thinking without penalizing the reasoning needed to reach the answer. Experiments across diverse tasks show the effectiveness of DRE.

Compositional learning systems must balance plasticity, the ability to acquire new knowledge, with stability, the preservation of previously learned components, especially when tasks share structure and risk interference. We study how modular architecture, task similarity, and representational dimensionality jointly shape compositional continual learning in a sequential A-B-A paradigm, comparing a task-partitioned recurrent network to a single-network baseline while inducing high- and low-dimensional regimes via weight-scale manipulations. In a high-dimensional "lazy" regime, both architectures achieve similar performance and internal geometry, suggesting that explicit modular structure has little impact when representations are weakly constrained. In a lower-dimensional "rich" regime, modularity becomes decisive: the modular network develops graded task-specific subspaces that overlap for similar tasks, partially align for moderately dissimilar tasks, and separate for dissimilar tasks, yielding a more compositional and interpretable organization than the single network. These findings identify the representational regime induced by initialization scale, which co-varies with representational dimensionality, as a key factor governing when compositional, modular structure is functionally beneficial in continual learning, and support viewing safety and robustness as problems of adaptive allocation of representational subspaces rather than fixed separation versus sharing.

Monotonicity has been a long-running architectural inductive bias for neural networks, motivated by tabular, scientific, and economic settings where outputs are known to respond monotonically to certain inputs. Existing approaches are MLP- or flow-based and lack per-edge functional transparency; the only Kolmogorov--Arnold Network (KAN) variant with monotonicity, MonoKAN, enforces the constraint only on a restricted parameter subset and requires a projection-style training procedure. We close this gap with \textbf{MKAN}, a KAN with hard monotonicity guaranteed for \emph{all} parameter values via exponential reparameterization of B-spline coefficients, positive edge weights, and a monotone base activation. Training reduces to standard unconstrained gradient descent. Our headline theoretical contribution is a \emph{representation-cost} theorem: any $C^K, K >0$ feature extractor inducing a ball-shaped semantic-neighborhood partition admits a monotone realization of the equivalent neighborhood structure at $N' = N^* + k \le 2N^*$, where $k$ is the number of non-monotone coordinates of the original. The bound is architecture-agnostic and gives a principled sizing rule for monotone encoders. Empirically, MKAN is competitive with state-of-the-art monotone NNs on the SMM/ICML-2024 benchmark while being the only method that combines hard unconstrained monotonicity with KAN's per-edge functional transparency; the $2N^*$ prediction is validated in a self-supervised feature-size sweep on four real datasets, and on a controlled monotone-generative dataset MKAN recovers ground-truth factors with substantially higher Spearman alignment than KAN, MLP, and linear baselines.

Large language models (LLMs) outperform earlier architectures on generative inference and long-context tasks, but their large size introduces significant challenges in memory usage, energy cost, and on-device deployment. Since scaling pre-trained language models improves downstream capability \cite{zhao2023survey}, the key-value (KV) cache becomes a dominant inference bottleneck. Recent KV cache compression methods \cite{jo2025fastkv,li2024snapkv,zhou2024dynamickv} reduce this cost by retaining only a subset of attention-relevant tokens. However, while these approaches preserve accuracy on benign workloads, their compression policies either fail to defend against jailbreak attacks \cite{jiang2024robustkv} or degrade safety alignment under aggressive eviction. We propose AnchorKV, a drop-in modification to KV cache compression that biases token retention scores away from directions in key space associated with harmful prompts. AnchorKV constructs an offline safety anchor by adapting a difference-of-means representation engineering approach \cite{arditi2024refusal,zou2023representation} to the layer-specific key projection space used in KV caching. Based on this anchor, a soft penalty token selection rule trades a small amount of utility for substantially improved safety alignment, while reducing to the original compressor when the penalty is zero.

Web navigation requires agents to follow natural language goals, interact with web pages, and produce accurate answers. While recent advances leverage vision-language models and reinforcement learning, existing methods still suffer from single-step fragility due to reward misalignment and error propagation. To tackle the reward entanglement, we design Dynamic Dual-Policy Optimization (DDPO), which dynamically switches between a navigation-first mode for exploration and an answer-first mode for question-answering to mitigate reward conflict. To calibrate the single-step error, we propose Confidence-Guided Adaptive Navigation Reflection (CANR), a mechanism that estimates per-step confidence, triggers reflection only when necessary, and uses contrastive rewards to encourage self-correction to calibrate the single-step inaccuracy. With the above as the main components, we finally develop our StepGuard, a new framework of Guarding Web Navigation via Single-Step Calibration. Experiments demonstrate that our approach significantly improves navigation and answer accuracy, setting new state-of-the-art performance on standard web navigation benchmarks.

Game generation is an emerging application of coding agents, requiring models to transform natural-language specifications into playable interactive systems. Unlike traditional coding tasks, game generation takes place within a game engine, where scripts, scenes, assets, rendering, and runtime interactions must jointly produce coherent gameplay. We formalize end-to-end game generation as the problem of producing a complete game artifact that realizes a specification through observable player-game interaction in a target environment. We argue that evaluating this setting requires three desiderata: Engine Grounding, Artifact Completeness, and Interactive Verification. We propose an interaction-grounded evaluation framework that assesses executable gameplay through replayed demonstrations and rubric-guided multimodal judging. We instantiate this framework as GameCraft-Bench, a benchmark comprising 140 Godot tasks across 15 game families. Evaluations of frontier coding agents show that end-to-end game generation remains highly challenging: the strongest agent achieves only 41.46%, and most agents score below 40%. Further analysis reveals that while agents often implement recognizable mechanics, they struggle to deliver complete games with sufficient content, functional visual feedback, and coherent presentation. See https://tongxuluo.github.io/gamecraft-bench-website for demos, code, and data.

Graph-based retrieval-augmented generation (GraphRAG) is effective for knowledge-intensive and multi-hop query tasks; however, many existing methods primarily seed entity-based graphs and rely on implicit semantic relevance propagation. This often (i) under-retrieves when user queries are abstract and semantically sparse at the entity level, and (ii) suffers from brittle multi-hop reasoning, where noisy activations can derail entity-to-entity transitions and corrupt the inferred relation chain, yielding unreliable conclusions. To this end, we propose \texttt{FlowRAG}, a semantic-aware retrieval framework that improves both semantic recall and explicit reasoning. Specifically, \texttt{FlowRAG} constructs a quad-level heterogeneous graph over passages, summaries, sentences, and entities, where summary nodes serve as a coarse semantic hub. At retrieval time, a dual-granularity activation module combines summary--query alignment with sentence-level matching to activate relevant entities under paraphrase and abstraction robustly. We then introduce a frequency-aware weighted flow module that routes relevance through entity--passage links weighted by within-passage term frequency, pruning noisy connections and extracting high-confidence reasoning paths as an explicit logic skeleton for generation. Extensive experiments show that \texttt{FlowRAG} obtains state-of-the-art performance on complex reasoning benchmarks.

A wide range of neurosymbolic (NeSy) systems compute one functional: a belief-weighted sum of a logical quantity over a space of $σ$-structures, of which weighted model counting, fuzzy logic, and probabilistic logic are special cases. This account is built on sets, and a set deliberately forgets two things that are important for NeSy: when two $σ$-structures are the same up to a symmetry of the theory, and how many distinct proofs witness a query. Replacing the underlying sets by types, in the sense of homotopy type theory, preserves this information, and turns this functional into a belief-weighted homotopy cardinality, a notion of size that counts each object in inverse proportion to its symmetries. We develop the framework from scratch for NeSy systems, prove a conservativity theorem that recovers the classical functional when symmetries are trivial, and show that the symmetry our framework exposes is exactly the one behind reasoning shortcuts. The payoff is concrete: the shortcut-aware concept posterior that recent methods reach by ensembling or expressive density estimation is the only symmetry-invariant point of the confusion-set simplex, computable in closed form by averaging a single model over the symmetry group. On MNIST reasoning-shortcut benchmarks this single-model wrapper is better calibrated than a diversity-trained ensemble, while leaving label accuracy and identifiable concepts untouched. Code is freely available at https://github.com/bio-ontology-research-group/hott-nesy.

WallGo is a recently introduced strategic board game popularized by the 2025 Netflix series The Devil's Plan. Although played on a small 7 x 7 board, its combination of stone movement and wall placement yields high game-tree complexity and intricate strategic interactions. Despite its growing popularity, WallGo remains underexplored. This paper presents WallZero, an AlphaZero-based agent for the two-player WallGo setting. We introduce tailored action and feature designs to improve playing performance significantly. In the evaluation, WallZero defeats two professional Go players who participated in this study, securing on average 1.98x more territory per game. Beyond its strength, we use WallZero to assess game fairness and identify key strategies for mastering WallGo. Interestingly, our results show that the opening used in the Netflix series yields a more balanced game. Our code is available at https://rlg.iis.sinica.edu.tw/papers/wallzero.

Foundation models in language and multimodality achieve strong generalization by aligning heterogeneous data under a unified formulation and training at scale. In this report, we investigate whether this scaling recipe can be applied to robotic manipulation to achieve genuine generalization. This is challenging because, unlike text, manipulation data is heterogeneous by nature, expensive to collect, and narrow in diversity, making alignment and scale simultaneously difficult. We present Qwen-RobotManip, a generalizable Vision-Language-Action foundation model built on Qwen-VL. Qwen-RobotManip introduces a unified alignment framework across the representation, motion, and behavioral dimensions of manipulation, making large-scale multi-source training coherent rather than conflicting. This alignment capability in turn enables Qwen-RobotManip to absorb manipulation data at a scale that prior training regimes could not sustain. A human-to-robot synthesis pipeline converts egocentric hand demonstrations into robot trajectories across 15 platforms, and a rigorous curation pipeline harmonizes heterogeneous datasets. Using only open-source datasets and human videos without proprietary data collection, Qwen-RobotManip constructs a ~38,100-hour pretraining corpus and exhibits emergent generalization capabilities, including zero-shot instruction following, robustness to perturbations, reactive error recovery, and cross-embodiment transfer. We find that standard benchmarks fail to capture pretraining quality and instead adopt OOD settings including RoboCasa365, LIBERO-Plus, EBench, RoboTwin-Clean2Rand, RoboTwin-IF, and RoboTwin-XE. Qwen-RobotManip substantially outperforms prior state-of-the-art models, including $π$0.5, across all OOD settings, ranks 1st in RoboChallenge with a 20% relative improvement, and is validated on real-robot platforms including AgileX ALOHA, Franka, UR, and ARX.

Drones hold promise for supporting emergency services, but their integration into workflows remains ad hoc and coordination-intensive. This paper addresses two research questions: how emergency teams want to collaborate with drones, and how to formalize these collaborations into repeatable processes. Based on four field trials and 95 interviews, we derive 44 interaction patterns grouped into 10 meta-patterns reflecting operational needs such as reconnaissance, communication, and logistical support. To structure these practices, we introduce DroneLets - a new class of design artifacts that extend Collaboration Engineering to embodied agents. DroneLets capture setup requirements, drone capabilities, environmental constraints, and coordinated actions across human and drone actors. They offer a modular framework for designing repeatable, scalable collaboration processes in emergency services, illustrated through patterns such as broadcasting to bystanders and post-fire monitoring. This work expands the scope of CE and provides a structured foundation for integrating autonomous drones into high-stakes field operations.

LLM agents in interactive environments are highly sensitive to their prompts, yet prompt engineering remains a manual, task-specific process. We introduce an automated prompt optimization framework for LLM agents that decomposes the observation-to-action pipeline into a goal-conditioned descriptor agent and an action selection agent, and iteratively refines each module's prompt through an LLM-driven evolutionary loop guided by environment returns. We propose a behavior analyzer to attribute episode outcomes to specific prompt components, and a mutator to propose targeted revisions to the prompt, before validating them through environment rollouts. We evaluate on all five BabyAI tasks in the BALROG benchmark, comparing our pipeline against BALROG's RobustCoTAgent under both plain and guided prompt initializations. Optimization improves performance consistently across tasks and conditions, without requiring updates to the model weights. On PutNext, a multi-step coordination task where the RobustCoTAgent achieves 0% success, our framework reaches up to 72.5% success rate using the same underlying LLM with optimized prompts. These results suggest that a multi-agent framework, combined with automatic prompt optimization, enhances LLMs without the need for fine-tuning or extensive human supervision.

Humanoid robots promise whole-body interaction in human-centered environments, but scalable policy learning remains difficult because task-level decision-making and whole-body dynamic execution are tightly coupled. A practical solution is hierarchical control, where a high-level policy predicts intermediate whole-body actions and low-level general motion trackers (GMTs) execute them as stable humanoid motion. However, existing benchmarks rarely evaluate the policy-tracker interface itself, leaving open whether intermediate whole-body actions are executable, robust under task distribution shifts, and transferable across different GMT backends. We introduce HumanoidArena, a simulation-first benchmark for egocentric hierarchical whole-body learning. The benchmark formulates policy learning as a hierarchical decision making problem: a high-level policy converts egocentric vision, proprioception, and instructions into a compact whole-body action, which is subsequently executed by a low-level GMT. Instead of treating the legs as planar transport tools, HumanoidArena emphasizes interactions where lower-body coordination is structurally necessary in task completion. We therefore design 7 leg-critical HOI/HSI tasks in which success requires foot placement, balance maintenance, posture adjustment, and whole-body reorientation. To further diagnose the hierarchical system, we evaluate policies from two complementary perspectives: perturbation-conditioned generalization and GMT-conditioned transfer. Experiments show that hierarchical control enables learned policies to solve diverse leg-critical interactions, but performance is strongly tracker-conditioned and cross-GMT transfer remains fragile. These results position HumanoidArena as a benchmark for studying transferable intermediate action representations and scalable egocentric whole-body policy learning.

Large language models (LLMs) are often hypothesized to perform implicit Bayesian inference, yet a key coherence condition, the martingale property of predictive beliefs, has been shown to fail in controlled synthetic in-context learning settings. We revisit this question in a more typical usage regime: generic multiple-choice question answering. Exploiting the discrete answer space, we compute exact predictive distributions and study belief dynamics induced by autoregressive answer resampling. We introduce prompted predictive resampling (PPR), where an LLM generates a sequence of answers to the same question. Empirically, PPR reveals early-stage belief drift, indicating martingale violations. However, after sufficient resampling steps, the belief process self-stabilizes and converges to a coherent predictive distribution. Based on this observation, we further propose (i) a seed-answer prompting strategy to accelerate stabilization, and (ii) a self-consistency loss that amortizes early-stage drift into the model via fine-tuning. Experiments on multiple-choice QA benchmarks show that our methods substantially reduce belief drift and improve predictive coherence without sacrificing accuracy.

Automatic speech recognition (ASR) in non-English clinical settings is challenged by multiscript variability, where the same term may appear in multiple valid orthographic forms. Conventional string-matching evaluation metrics often underestimate ASR performance by treating orthographic variants as errors. To address this issue, we introduce MultiClin, a clinical ASR benchmark designed to evaluate robustness to multiscript variability. Experiments across diverse ASR models show that multiscript-aware evaluation provides a fairer assessment of recognition quality than conventional single-reference evaluation. We further investigate the impact of script consistency during training and find that inconsistent script mappings increase orthographic uncertainty and hinder model convergence, with a balanced 50% mapping ratio producing the highest entropy. In contrast, script unification consistently yields the best ASR performance. Our dataset and code are publicly available at: https://github.com/aitrics-ronaldo/Interspeech_MultiClin.

Segmenting 3D assets into meaningful regions remains challenging, especially when segmentation criteria are application-dependent and require user control. We present a human-in-the-loop pipeline for generating a segmented 2D parameterized atlas from a 3D model for interactive media, game, and XR content workflows. Our method first selects a compact set of rendered views using a greedy set cover strategy over sampled surface points, and then supports interactive segmentation of these views with SAM~2 and Label Studio. The resulting masks are back-projected onto the model's UV parameterization to produce a unified segmented atlas that supports downstream production tasks such as segment-wise material assignment, style transfer, and semantic labeling. We assess the pipeline through a demonstration-based technical evaluation on eight cultural heritage objects. The results show that the approach can generate usable segmented atlases across diverse geometries while revealing recurring sources of manual correction, particularly fine structures, cavities, and weak appearance boundaries.

Large Language Models (LLMs) have demonstrated remarkable capabilities in translating natural language to SQL, yet existing methods still falter on complex queries requiring multi-step, data-aware reasoning. We introduce DecoSearch, a training-free framework that addresses this by routing each query to the appropriate level of reasoning effort. A lightweight Schema Selector first prunes the full database schema to the relevant tables and columns. An LLM Judger then decides whether the question requires decomposition: straightforward questions follow a direct generation path and complex ones are escalated to a Directed Acyclic Graph (DAG) of atomic sub-questions, each solved by a targeted SQL generation step. A RAG component grounds the decomposer with semantically similar training examples, and a Topology Refiner restructures the reasoning plan when execution failures signal a flawed decomposition rather than a fixable SQL error. DecoSearch achieves 70.53% execution accuracy on BIRD and 88.31% on Spider with a DeepSeek backbone, surpassing all training-free baselines while consuming an order of magnitude fewer tokens than competing methods. It also functions as a model-agnostic wrapper, consistently improving fine-tuned SQL generation backbones without any modification to the pipeline.

Agent skills -- structured, reusable knowledge artifacts that augment LLM agent capabilities -- have been rapidly adopted in industry, yet their cross-domain impact and use across commercial and open-source models remain under-studied, and no reusable methodology exists for evaluating an individual skill. In this work, we present an evaluation framework that lets a skill author construct realistic tasks to rigorously assess the aspects of a skill that matter most to them, and that estimates skill utility by solving those tasks. Further, we apply our evaluation approach at scale to 500 real-world skills, generating 1,000 tasks derived from the skills' content, along with instruction-following and goal-completion scoring rubrics. Using these metrics, we evaluate how 19 agent-model configurations, both proprietary and open-source, perform on the tasks. Our results show that models vary widely in how closely they adhere to the instructions encoded in skills, leading to substantial differences in their performance gains. Furthermore, we show that access to a skill significantly changes model behavior compared to the no-skill setup, providing an essential mechanism for encoding opinionated workflows into LLM agents. We release our evaluation dataset to support future work on agent skills.

Contrastive Language-Image Pre-training models are widely reused across downstream interfaces, including feature extraction, retrieval, reranking, and selection. Existing CLIP backdoor, however, usually validate attacks on a small attack-native task, leaving unclear whether the same poisoned checkpoint remains exposed, weakens, or becomes not applicable when reused through other interfaces. We introduce DIFE, a Deployment-Interface Footprint Evaluation framework that audits backdoored CLIP checkpoints across deployment interfaces. DIFE makes various evaluations comparable by specifying each interface's component readout, trigger channel, target event, reference condition, and metric. DIFE also introduces effective-footprint diagnosis to identify the reusable CLIP component or component combination that carries exposure and explains where risk transfers. Auditing reproduced CLIP backdoors with DIFE reveals a structured landscape: native success is not a checkpoint-level risk certificate, exposure follows component footprints, text-side poisoning does not yield textual-encoder control, and some coupled attacks remain mechanism-bound. This audit reveals a import gapin existing CLIP backdoors: a textual encoder that itself becomes a reusable carrier of adversarial behavior. We therefore introduce BadTextTower to fill this gap. BadTextTower produces strong text-conditioned retrieval, reranking, and selection exposure while leaving visual-only reuse nearly clean.

The dynamic Leidenfrost effect LFE and behaviour of impinging colloidal droplets is strongly influenced by the impact and spreading paradigms. LFE actuated rebound and levitation occurs due to enhanced spreading and near-frictionless recoil over the intervening vapour layer, providing opportunities for external field stimulus aided modulation and control of impact outcomes, and the resulting boiling-LFE behaviour. Magnetic field modulated LFE onset, dynamics and boiling transport of stable aqueous nano Fe2O3 based ferrofluid droplets was studied using high speed imaging. The interplay between magnetic, inertia, and viscocapillary forces on droplet spreading, magneto LFE-driven rebound conditions, residence time, and post-impact regimes was analysed using dimensionless parameters maximum spread factor, Weber number, and magnetic Bond number. We report a purely new phenomenon, namely magneto Leidenfrost effect MLFE, wherein magnetic field induces LFE aided onset of droplet rebound at substrate temperatures Ts below the zero-field dynamic Leidenfrost temperature LFT. The critical for the onset of MLFE decreases with increasing . Increasing the nanoparticle concentration permits the onset even at considerably lower . At elevated Ts , the residence time is noted as dependent. At much higher Ts, increasing promotes formation of radial filamentous structures, leading to complete droplet fragmentation. We also propose a theoretical framework that explains magnetic field driven spreading enhancement and rebound, and predicts of MLFE droplets in agreement with experiments. Our findings provide valuable insights into the novel realm of field dictated LFE, and hold significant implications towards the design of frictionless, rapid colloid droplet transport systems, and targeted droplet manipulation or activation for advanced thermal management.

Automated pollen identification from microscopy remains a bottleneck in aerobiology, palaeoecology and biodiversity monitoring, because scalable systems must generalise across specimen preparation, scanner settings and geographic origins while retaining palynological interpretability. To address this gap, we present a million-scale multimodal pollen microscopy resource, Pollen AI Atlas, assembled from pure-species whole-slide bright-field images spanning four geographic origins, four scanner settings and 46 taxon labels across 31 botanical families. Seeded by one manually selected exemplar per source slide, token-level mining and filtering produced 1,511,390 released grain detections with 99.6\% proposal precision in expert-curated test regions. Each detection was paired with machine-generated grain-level morphological captions from five open-weight vision-language models, guided by expert-verified palynological anchors, yielding structured descriptions of aperture systems, wall ornamentation, shape and size. Among the evaluated models, Gemma4 provided the most controlled primary caption set, combining tight length control, no leakage and the strongest text-retrieval performance. Baseline benchmarks with frozen visual features reached 88.16\% top-1 accuracy, while cross-regional retrieval showed that caption-derived text embeddings remained robust when image similarity degraded (mAP@20 0.811 versus 0.262). Released data, annotations, captions, splits, code, and weights provide a benchmark for pollen recognition, cross-regional domain adaptation and domain-specific multimodal microscopy learning.

Large language models achieve strong reasoning performance by scaling inference-time compute, yet remain fundamentally stateless, discarding the rich, self-produced reasoning traces generated during this process. We investigate whether models can instead learn online from this experience, converting transient computation (reasoning traces) into persistent reusable knowledge, and without external supervision or access to future data. We show that In-Context Learning (ICL) over raw reasoning traces fails to generalize, reflecting a fundamental limitation of token-level reuse: individual traces lack the abstraction needed for transfer, even after refinement (e.g. self-reflection). In contrast, drawing inspiration from recent works on unsupervised reinforcement learning, we find that lightweight per-instance training with self-generated test-time signals (majority voting) as rewards yields substantial gains, often surpassing full-dataset offline training, motivating a shift from raw traces to learned latent representations. Building on this insight, we propose an online method that distills inference-time compute spent on encountered problems into compact modular latent memories capturing the underlying reasoning structure. These memories are stored and retrieved for future inputs, enabling continual improvement while avoiding catastrophic forgetting through modular design. Importantly, our method is highly efficient, parametrized as extremely lightweight soft prompt memories (~0.001% of model parameters) and trained with only a few gradient steps, yet achieving performance competitive with full parametric updates and offline training. Across challenging mathematical reasoning benchmarks, our approach significantly outperforms zero-shot and raw data ICL baselines, while transferring effectively across datasets.

As an increasing majority of global video content is consumed on social platforms for interactive social purposes, video generation models built for social worlds are important but largely overlooked by previous studies. In this work, we define the position of social world models and build a prototype model as the first step towards this goal. While previous world models successfully simulate physical environments or gaming world exploration, they remain fundamentally detached from human-centric social dynamics. To bridge this gap as the first step to social world models, we present MaineCoon, the first real-time audio-visual autoregressive model that has 22B parameters and is capable of real-time streaming generation and sub-second interaction, with a record-breaking frame rate of up to 47.5 FPS, on a single GPU. To the best of our knowledge, MaineCoon is also the first real-time audio-visual generation model specifically optimized for social-interactive applications. To enable efficient and stable training, we introduce several novel techniques into MaineCoon, including self-resampling, cross-modal representation alignment, domain-aware preference optimization, and reinforced online-policy distillation (ROPD). We also design the first agentic streaming inference framework that supports thousand-second-scale or even longer generation while mitigating drift with agentic cache management and prompt planing. These innovations significantly accelerate training while optimizing real-time inference performance. We believe this work not only sets a new state-of-the-art (SOTA) performance benchmark for high-quality, low-latency, and long-horizon audio-visual autoregressive models, but also points out the paradigm shift desired for next-generation AI-native social platforms.

Coding agents have become a major mode of software engineering, but the benchmarks we use to compare them were designed in a pre-agent era: they collapse model, harness, and environment into a single end-to-end score, typically computed against one reference solution, with no component-level signal for iteration. We argue that current coding benchmarks are misaligned with agentic software engineering. A coding agent in practice is not a model: it is a system harness -- a composite of models, harnesses, contexts, environments, and feedback signals, any one of which can move the benchmark score by margins comparable to those between adjacent model generations. We discuss three symptoms: (i) benchmark scores conflate the model with the rest of the harness; (ii) grading against a single reference solution penalises equally valid alternatives; and (iii) the absence of signal at the level of individual harness components makes the end-to-end system score difficult to iterate on.

Despite the remarkable progress of Video Large Language Models (Video-LLMs), current online architectures still struggle to simultaneously process continuous video streams, decide autonomously when to respond, and preserve long-horizon contextual memory. These obstacles undermine real-time responsiveness and cause severe forgetting throughout prolonged interactions. In this work, we introduce LiveStarPro, a live streaming assistant that is designed for proactive video understanding over long-horizon streams. The design of LiveStarPro rests on three complementary components. The first component is Streaming Verification Decoding (SVeD), an inference framework that identifies the appropriate response timing through single-pass perplexity verification, thereby eliminating the dependency on explicit silence tokens. The second component is Streaming Causal Attention Masks (SCAM), a training strategy that enforces incremental video-language alignment over variable-length streams. The third component is Tree-Structured Hierarchical Memory (TSHM), a recursive memory architecture that organizes evicted historical information into event chains and consequently enables efficient retrieval from effectively unbounded video streams. To facilitate a comprehensive evaluation under realistic online conditions, we further present OmniStarPro, a large-scale benchmark that spans 15 diverse real-world scenarios and that extends to hour-scale streams for the assessment of long-term recall. Extensive experiments demonstrate that LiveStarPro consistently surpasses existing methods, attaining a 28.9% improvement in semantic correctness and an 18.2% reduction in timing error, while its streaming key-value cache further yields a 1.58x inference speedup over the same model without caching. The model and the code are publicly available at https://github.com/sotayang/LiveStarPro.

Training psychotherapists in evidence-based interventions such as Acceptance and Commitment Therapy (ACT) requires repeated practice with meaningful feedback, yet opportunities for safe, standardized training are limited by ethical, logistical, and resource constraints. We introduce a system designed to support ACT-oriented psychotherapy training through spoken dialogue with an embodied virtual patient. The system uses large language models to simulate patient behavior conditioned on profiles derived from real therapy sessions and configurable clinical scenarios, while a separate automated evaluator provides turn-by-turn feedback on therapist responses based on established ACT fidelity criteria. Rather than aiming to replace supervision, the system is intended to support deliberate practice by enabling experimentation, reflection, and immediate feedback in low-risk settings. Expert evaluation with practicing psychologists confirmed high realism in patient behavior and demonstrated that immediate turn-by-turn ACT feedback increased therapists' awareness of intervention choices and enabled effective experimentation with alternative responses. Quantitative evaluation across 49 therapy transcripts identified GPT-4o-mini as the optimal feedback model, achieving the lowest mean absolute error (MAE = 6.12) in replicating human supervisor ACT fidelity ratings with statistically significant agreement. This work demonstrates the potential of fidelity-aware simulated patients as a scalable complement to psychotherapy training.

Personal health data from wearables are typically presented through dashboards of charts and summary statistics, requiring users to actively interpret patterns and implications. We explore an alternative interaction paradigm: engaging with personal health data through an embodied conversational agent that facilitates objective data reflection in dialogue with the user. We present a system that combines lightweight preprocessing of wearable data with a Unity-based embodied character. Internally, the system follows a dual-agent design in which an Observer agent extracts descriptive statistics and temporal trends, and a Presenter agent communicates these findings through "spoken statistics," intentionally refraining from clinical advice to isolate the impact of the interaction modality. We evaluate this approach through a simulated-self user study (N=5) using a within-subject design. Participants adopted health personas and goals derived from the LifeSnaps dataset to compare traditional dashboard exploration with embodied conversational reflection. Our evaluation focuses on perceived understanding, the specificity of generated actions, and the cognitive shift from passive viewing to active sensemaking. The paper contributes a functional prototype, a design pattern for objective health data narrative generation, and early empirical insights into how embodiment affects the interpretation of personal health metrics.

A-mode ultrasound (US) is a promising sensing modality for Virtual Reality (VR) interaction, as it enables the mapping of muscular activity into control commands while retaining the benefits of wearable sensing. However, existing approaches still face limitations in terms of wearability and interaction complexity, often relying on external hardware such as cameras. In this work, we propose a fully wearable multimodal interface for real-time VR-interaction, based on concurrent US and inertial (accelerometry) sensing from the forearm and upper arm. The system is built on the WULPUS platform and integrates an end-to-end software framework for real-time acquisition, visualization, and communication with a Unity-based VR environment. A multimodal learning pipeline is introduced for concurrent hand pose and forearm position estimation in 2D space. The interface is evaluated through offline and online experiments with five subjects, during the execution of three functional tasks: cylinder grasping (gross motor) and relocation, marble pinching (fine motor) and relocation, and liquid pouring. For offline experiments, we collect 5 acquisition sessions across multiple days, achieving an average inter-session accuracy across subjects of 80$\pm$6\% for hand pose estimation and 77$\pm$7\% for forearm position estimation. Online validation with minimal fine-tuning (5 min) demonstrates success rates of 92.0$\pm$16.0\%, 88.0$\pm$9.8\%, and 96.0$\pm$8.0\% for the three tasks, respectively. With a power consumption of only 19.9~mW, our system enables more than 2.5 days of continuous use on a small 350 mAh LiPo battery without the need for recharge, enabling truly wearable, multimodal, and functionally meaningful VR interaction.

Although reinforcement learning (RL) has expanded the cognitive boundaries of large language models (LLMs), it often remains vulnerable to the autoregressive curse in long-horizon logical reasoning: small epistemic perturbations introduced early in generation can propagate irreversibly along the Markov decision process flow, triggering cascading failures that drive the reasoning trajectory toward collapse. To overcome this autoregressive cascade, in which a single early mistake can compromise all subsequent reasoning steps, we propose dynamic epistemic entropy orchestrated erasable reinforcement learning ($\text{E}^3\text{RL}$). $\text{E}^3\text{RL}$ eliminates reliance on external signals by grounding the model's endogenous local autoregressive cross-entropy as an intrinsic coordinate of epistemic uncertainty. By introducing segment-level adaptive dynamic thresholds and advantage allocation, $\text{E}^3\text{RL}$ enables the model to precisely excise localized logical defects while reusing historical key-value (KV) cache streams, thereby endowing the reasoning process with a self-healing capability. We train $\text{E}^3\text{RL}$ on the DeepMath-103k dataset. Experimental results show that $\text{E}^3\text{RL}$ reshapes the exploration efficiency of long-sequence reasoning and improves sample efficiency while maintaining linear memory overhead. On mathematical reasoning benchmarks such as AIME, $\text{E}^3\text{RL}$ achieves substantial performance gains, with the 4B and 8B parameter models surpassing previous state-of-the-art (SOTA) results by 5.349\% and 6.514\%, respectively. These findings suggest that $\text{E}^3\text{RL}$ shatters the autoregressive curse in long-sequence reasoning and establishes a theoretical and systems-level foundation for the next generation of self-healing artificial general intelligence (AGI).

Interactive world models aim to simulate environment dynamics under real-time user actions. However, their action vocabulary is largely confined to navigation: most actions correspond to motion (e.g., walk, turn, look around), while interaction with objects in the scene (e.g., pick up plates, open doors, or trigger physical responses) is either absent, restricted to game domains, or relegated to prompt-to-full-video scenarios. The resulting worlds are visually explorable but not truly actionable. In this work, we present ActWorld, an interactive world model that extends prior navigation-centric generators to support mid-rollout object interaction within a chunk-autoregressive framework. We argue that the navigation-interaction gap stems from two bottlenecks. First, a data bottleneck: the lack of human-object interaction data with accurate, dense labels. Second, a memory bottleneck: recency-biased history compression in existing world models discards the event-transition frames that causally determine subsequent object states, leading to an action-forgetting pathology. On the data side, we construct a 100K interaction video dataset, each annotated with per-chunk captions via chain-of-thought reasoning. On the model side, we introduce a hierarchical action-aware memory design that routes history compression by interaction importance, complemented by a persistent memory bank that maintains event-update and object-identity tokens across long rollouts. Experiments show that ActWorld supports both flexible navigation and rich object interaction within a single model, substantially improving interaction fidelity over navigation-only baselines without sacrificing viewpoint control. Project page is available at https://interactwm.github.io/ActWorld.

Recent vision-language models (VLMs) have shown promising progress in generating webpages from visual inputs, yet existing evaluations mainly focus on short, single-screen, and largely static webpages. We introduce LongWebBench, a benchmark for evaluating long-horizon webpage generation from both structural and functional perspectives. LongWebBench contains 490 real-world long webpages for structural fidelity evaluation and 507 goal-oriented interaction tasks over 129 webpages for functional evaluation. It employs two complementary protocols: a multi-dimensional VLM-based metric for assessing long-range structural coherence, and a DOM-augmented agent-based pipeline for end-to-end functional verification. We further examine the automatic evaluation protocols through human agreement analysis. Experiments with state-of-the-art open-source and proprietary VLMs under single-image and multi-image settings reveal that structural fidelity degrades as webpage length increases, while visually plausible generations often fail to support executable multi-step interactions. These results highlight the need to evaluate long webpage generation beyond visual similarity, with executable interaction as a core criterion. Our code and data are available at https://github.com/zheny2751-dotcom/LongWebBench.

Optical remote sensing imagery is frequently degraded by cloud and cloud-shadow contamination, which limits its reliability for near-real-time land use and land cover (LULC) mapping. Although synthetic aperture radar (SAR) can provide cloud-penetrating structural information, existing SAR-optical fusion methods often assume reliable optical observations and insufficiently address the semantic uncertainty introduced by cloud contamination. To address this issue, we propose CloudLULC-Net, an end-to-end heterogeneous SAR-optical fusion framework that directly predicts LULC maps from cloud-contaminated Sentinel-2 imagery and temporally adjacent Sentinel-1 SAR observations. The proposed network incorporates optical reliability modulation to suppress unreliable optical responses, heterogeneous information adaptive aggregation to model high-order spatial-channel interactions between optical and SAR representations, and a unified semantic mapping transformer to organize fused features in a LULC-oriented latent space. A semantic anchor-guided optimization strategy is further introduced to improve the consistency of intermediate semantic representations. To support this task, we construct CloudLULC-Set, a large-scale benchmark dataset containing 40,223 curated SAR-optical-label triplets with pixel-level LULC annotations across diverse geographic regions and cloud conditions. Experimental results show that CloudLULC-Net achieves an OA of 86.60%, an F1-score of 83.29%, and an mIoU of 73.51%, outperforming representative heterogeneous reconstruction-first and end-to-end SAR-optical mapping methods. Comparisons with existing global LULC products and analyses under different cloud-cover levels further demonstrate the robustness and practical value of CloudLULC-Net for target-date LULC mapping in cloud-prone regions.The project is publicly available at: https://github.com/RSIIPAC/CloudLULC

Medical vision-language models report strong chest radiograph accuracy, and this is increasingly read as evidence that they use the image. That inference is unsafe: a model exploiting finding-name priors scores like one that reads the scan, and no standard benchmark separates them. We introduce a causal audit that intervenes on the image, occluding the relevant region, occluding an irrelevant one, and swapping in another patient's same-label scan, and combines three behavioral metrics to test whether a correct answer depends on the image. Across nine systems, a text-only model with no image access reaches within 5.7 accuracy points of the best multimodal one, and a 119-billion-parameter multimodal model is statistically indistinguishable from a 7-billion text-only baseline. The audit splits the cohort into three models that ignore the image, one that is unstable, and five that use it selectively, for a subset of findings; the categories hold across a second dataset, resolution, and prompt phrasing. Against board-certified radiologists, a text-only model is statistically indistinguishable from a radiologist's accuracy while grounding at zero, whereas the image-using models ground at radiologist-comparable rates. Reported confidence flags ungrounded answers only when a model uses the image. Grounding audits, not accuracy, should gate clinical deployment.

Curriculum learning couples two design choices, how samples are scored by difficulty and how harder samples are paced into training, making it difficult to attribute observed gains to either component. We disentangle these factors with two evaluation protocols: stage-wise test subsets that validate scoring functions independently of curriculum training, and a baseline that applies the same pacing schedule to randomly ordered data. Within the Transfer Teacher framework (TTF), we use these protocols to evaluate a confusion-aware difficulty score that considers both correct-class confidence and the probability distribution over incorrect classes. On CIFAR-10 with ResNet-18 and VGG-16, the proposed score produces model-interpretable difficulty rankings that align with human intuition. However, at full data, neither curriculum nor anti-curriculum ordering improves accuracy over standard training, indicating that improving the scoring function alone is insufficient to overcome the known failure modes of curriculum learning in TTF. In contrast, We find that confusion-aware curriculum ordering result in consistent data-efficiency benefits, outperforming random ordering by up to 8.7% points at the 20% data regime, suggesting the potential of TTF as a data-efficient training method.

As LLM-based shopping agents enter production, existing benchmarks fail to capture how a shopper's requirements arrive: stated implicitly in the query, recorded in a profile, or revealed only when the right question is asked. Benchmarks that expose full intent upfront and grade only the final choice can neither pose this long-horizon challenge nor explain which requirement an agent missed. To address this gap, we introduce EComAgentBench, a benchmark of 662 tasks grounded in real Amazon products and reviews. Each task scatters these requirements across a visible query, a tool-gated profile, and scripted clarification; an agent must uncover hidden intent, verify candidates against attributes and review evidence, and commit to a single product within 100 tool calls. Moreover, typed, source-tagged rubrics grade every task, attributing each failure to a requirement and its source. Construction is automated yet reliable, with every answer fixed in code before any text is generated and every sample validated. Our evaluation of seven models reveals that even the strongest attains only 57.1% overall accuracy, and rubric satisfaction degrades from visible to hidden sources. Overall, we believe EComAgentBench will serve as a reproducible foundation for moving shopping agents from single-query search toward dependable assistance over long horizons.

The optical binding of gold nanoparticles has conventionally been explored within the Rayleigh limit using dipole approximations. But the field is increasingly focusing on the Mie regime for particles in the 100-500 nm range, where the dipole approximation is insufficient, and a complex landscape of multipolar resonances must be considered. This can be leveraged to engineer more complex forms of optical matter. To this end, we computationally study the optical binding force landscapes experienced by a pair of AuNPs using generalized multiparticle Mie theory. We calculate the total optical binding forces and mechanical trap stiffness values ($dF_i/di$) at the specific resonance wavelengths where the electric dipole, quadrupole, or octupole modes reach their respective scattering peaks and dominate the mechanical response. We demonstrate that the plasmonic mode symmetry greatly influences the spatial distribution of zero-force nodes and the rigidity of the optically bound dimer. By aligning these multipolar phenomena with standard experimental configurations, this work provides a mechanical framework for programmable metafluids and reconfigurable micromachines, bridging the gap between fundamental electrodynamics and reconfigurable nanomanipulation.

Accurate short-term electricity load forecasting is critical for the reliable and economic operation of modern power systems, under non-stationarity arising from weather variability, calendar effects, and evolving consumption patterns. While deep learning models such as LSTMs and Transformers show promising performance, most existing studies focus on direct absolute load prediction without explicitly addressing target non-stationarity. Motivated by classical time-series differencing techniques in ARIMA models, this paper investigates a delta-based target reformulation for short-term electricity load forecasting using deep learning. Instead of directly predicting absolute load values, the proposed formulation trains models to predict the change in load between consecutive time steps, with final forecasts reconstructed using the last observed load. This aims to stabilize the learning target and reduce forecasting difficulty. Using multi-year, hourly real-world electricity load data from India, augmented with meteorological variables from the NASA POWER project and calendar features, this study evaluates LSTM and Transformer models under both formulations, benchmarking them against LightGBM. Experiments are conducted for hour-ahead and day-ahead horizons, assessing performance via Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). Results show that delta-based reformulation consistently improves forecasting accuracy for hour-ahead prediction across all evaluated models, yielding MAPE reductions of over 50% compared to absolute formulations. For day-ahead forecasting, delta targets specifically benefit deep sequence models (LSTM and Transformer), while LightGBM remains competitive under the absolute formulation. These findings indicate that while delta reformulation is a powerful inductive bias for neural networks, its efficacy is model- and horizon-dependent.

Despite remarkable performance on complex tasks, Large Reasoning Models (LRMs) often generate excessively long Chain-of-Thoughts (CoT), inflating computational costs even for simple queries. Existing efforts to mitigate this inefficiency typically rely on discrete reasoning modes or fixed budget tiers, lacking a principled criterion of when reasoning is sufficient. In this work, we introduce Minimal Sufficient CoT (MSC), defined as the shortest prefix of a CoT trajectory which is adequate for producing the correct answer. We empirically show that MSC not only reduces reasoning tokens, but also improves accuracy across difficulty levels. Building on MSC, we propose Sufficiency-guided Continuous Adaptive Reasoning (SuCo), a two-stage training framework for autonomous reasoning control along a continuous spectrum. In stage 1, MSC-Aligned Fine-Tuning (MFT) constructs MSC data using problem-adaptive sufficiency thresholds that naturally scale with question difficulty, then fine-tunes the model to internalize concise yet sufficient reasoning patterns. In stage 2, Sufficiency-Aware Policy Optimization (SAPO) further optimizes the model through reinforcement learning with dynamic complexity tracking and sufficiency-aware rewards that penalize both over- and under-thinking. Extensive experiments across mathematics, code, and science benchmarks show that SuCo consistently achieves improvements in both accuracy and reasoning efficiency.

While large language models (LLMs) have greatly advanced the functional correctness of automated code translation systems, the runtime efficiency of translated programs has received comparatively little attention. With the waning of Moore's law, runtime efficiency has become increasingly important for program quality, alongside functional correctness. Our preliminary study reveals that LLM-translated programs often run slower than human-written ones, and this issue cannot be remedied through prompt engineering alone. Therefore, our work proposes SwiftTrans, a code translation framework comprising two key stages: (1) Multi-Perspective Exploration, where MpTranslator leverages parallel in-context learning (ICL) to generate diverse translation candidates; and (2) Difference-Aware Selection, where DiffSelector identifies the optimal candidate by explicitly comparing differences between translations. We further introduce Hierarchical Guidance for MpTranslator and Ordinal Guidance for DiffSelector, enabling LLMs to better adapt to these two core components. To support the evaluation of runtime efficiency in translated programs, we extend existing benchmarks, CodeNet and F2SBench, and introduce a new benchmark, SwiftBench. Experimental results across all three benchmarks show that SwiftTrans achieves consistent improvements in both correctness and runtime efficiency.

Reinforcement learning pipelines for Large Language Model (LLM) training often rely on manually redesigned environments between stages, requiring practitioners to heuristically infer which configuration will best improve the current policy. To automate this process, we propose the LLM-as-Environment-Engineer framework in which the current policy model analyzes failure trajectories together with contextual information and proposes modifications to the next-stage training environment configuration. We also introduce MAPF-FrozenLake, a controllable testbed whose generator exposes multi-dimensional environment configurations, making it suitable for studying and benchmarking environment redesign. On this testbed, we condition the environment engineer on structured summaries of policy behavior, failure cases, and environment statistics, from which it produces the configuration for the next training stage. With Qwen3-4B as the backbone, our framework achieves the strongest aggregate performance on our benchmarks, outperforming larger proprietary LLMs (e.g., GPT, Gemini) and fixed-environment training baselines. We further analyze which forms of context are most effective, finding that successful environment updates rely on failure evidence and preserve configurations that already work. Interestingly, the current RL checkpoint serves as a better environment engineer than the original base model, suggesting that policy learning improves the model's ability to diagnose its remaining weaknesses.

Reinforcement learning (RL) has emerged as a powerful paradigm for training Large Language Models (LLMs) as agents. However, conventional RL methods for long-horizon agentic tasks often struggle with sparse outcome rewards. Intuitively, this overlooks the rich environment dynamics information contained in rollout interaction trajectories. We argue that the interaction experience inherently serves as an implicit supervision signal, reveals the underlying transition mechanisms of the environment, and enables the agent to construct a more accurate internal model of the environment.. Therefore, in this work, we investigate how to leverage this additional signal to improve policy learning. Specifically, we propose EnvRL, a framework that incorporates environment dynamics learning into agentic RL via two auxiliary objectives: state prediction and inverse dynamics. By jointly optimizing with the primary RL objective, we encourage the agent to internalize environment dynamics from its own interaction experience. Extensive experiments on two long-horizon agentic benchmarks demonstrate that EnvRL achieves significant improvements on success-rates over RL-only baselines, e.g., when trained with GRPO, lifting Qwen-2.5-1.5B-Instruct from 72.8% to 77.4% on ALFWorld, and from 56.8% to 67.0% on WebShop.

Multimodal large language models (MLLMs) integrate strong text reasoning with visual inputs, yet their responses can be inconsistent with the underlying images, indicating ineffective utilization of visual evidence during inference. The prevailing training paradigm relies on large-scale caption-based pretraining for general alignment, followed by supervised fine-tuning and reinforcement learning to enable instruction following and complex reasoning. However, such pretraining provides only weak visual grounding: short, coarse captions bias models toward salient objects while neglecting fine-grained visual evidence. In this paper, we introduce Visual Evidence Pre-Alignment (VEPA), an intermediate stage between pretraining and post-training that explores a novel sufficiency-driven objective with Group Relative Policy Optimization (GRPO) to optimize question-conditioned visual evidence descriptions. Extensive experiments across diverse benchmarks show that our VEPA consistently enhances performance on visually demanding evaluations and complements standard supervised post-training. Further analyses show that the income stems from strengthened, transferable visual grounding, rather than from additional task-specific training.

Photonic reservoir computing offers a hardware-efficient route to processing temporal and sequential data, but delay-based implementations often rely heavily on temporal multiplexing, where long temporal masks are required to generate a sufficiently rich reservoir state. Here we show numerically that the spatial degrees of freedom of an active multicore fiber placed inside a delayed optical feedback loop can reduce this dependence on serial temporal encoding. The input signal is encoded by temporal and spatial masks, the pump distribution across the cores controls the reservoir operating point through the core-dependent effective gain and saturation energy, and the detected core intensities serve as readout features for a single trained linear layer. The system is modeled by linearly coupled nonlinear Schrödinger equations with saturable gain and solved using a split-step Fourier method. On the Mackey-Glass one-step-ahead prediction benchmark, a seven-core reservoir with equal temporal masks reduces the validation normalized root mean square error from 0.5956 for the single-core baseline to 0.0651 at a modulation rate of 40 GHz. At 1 GHz, spatial-only encoding reaches an error of 0.0323 using one temporal sample per symbol and no temporal mask. These results show that an active multicore fiber can provide both parallel readout channels and a tunable nonlinear transformation, offering a route to photonic reservoirs with reduced reliance on temporal multiplexing.

Molecular dynamics (MD) simulation is computationally demanding, particularly for large-scale systems requiring long-term analysis. Accurate forecast of the outcomes of a MD simulation is not only an attractive scientific challenge but also has substantial practical value. In this work, we developed a data-driven framework, termed ASTEROID (Advanced Spatiotemporal TransformER fOr Inferring Dynamics), that can directly predict multi-step atomic coordinates, avoiding conventional iterative integration. For this purpose, our ASTEROID reformulates MD trajectories as high-dimensional spatiotemporal sequences and integrates the Spatiotemporal Information (STI) Transformation equation into a Transformer architecture. The core innovation of ASTEROID lies in its ability to model multiscale spatiotemporal dependencies. In particular, for spatial dependencies, a local-global self-attention mechanism captures both short- and long-range interactions. For temporal dependencies, an encoder-decoder structure integrates global context with autoregressive forecasting. ASTEROID was evaluated on several quantum-mechanics derived molecular datasets. Our results indicate that ASTEROID achieved not only a higher level of accuracy in multi-step prediction than existing methods on various benchmarks, but also significantly reduced computational cost of conventional MD simulation. Moreover, the model supports iterative multi-step forecasting over an extended time scale. This work establishes a robust and generalizable data-driven paradigm for accelerating MD simulations.

The high cost of fine-tuning LLMs poses a significant economic barrier; pre-hoc performance prediction offers a critical solution to substantially reduce this expense. However, the theoretical limits of pre-hoc performance prediction remain unexplored. We formulate it as a stochastic estimation problem under information constraints, decomposing prediction risk into two components: an intrinsic limit (static data-model compatibility) and a reducible optimization variance. We prove that optimization variance admits a necessary lower bound on its decay rate, implying fundamental constraints on how quickly uncertainty dissipates, regardless of the predictor used. Based on these dynamics, we derive a budget-optimal probing principle and introduce a predictability phase diagram that organizes tasks into three distinct regimes: Static-Sufficient, Dynamic-Critical, and Noise-Dominant. Extensive experiments on synthetic and real-world benchmarks validate these theoretical regimes and demonstrate the efficiency of our probing strategy.

Large language model (LLM) web agents are usually deployed as tool callers: each turn, the model reads a fresh page observation and emits one structured tool action. When every action is a low-level primitive, horizons grow quickly and so do policy-facing LLM completions, dominating latency and cost on benchmarks such as Mind2Web and WebArena. Recent systems therefore wrap repeated interaction fragments as web skills: callable tools built from successful trajectories or induced programs, so one call can replace several primitives. However, prior skill libraries are still triggered mainly by instruction similarity or coarse site metadata, which yields low skill reuse on held-out sites and leaves much of the potential step and token reduction on the table. We present SkillMigrator, an agent that learns reusable web skills and transfers them across sites by matching layout structure rather than specific element references. Each induced skill is stored as a transferable interaction pattern (TIP): the skill paired with a structural sketch of the snapshot at induction time. At test time, SkillMigrator retrieves TIPs by layout similarity and grounds their references on the live page. The rest of the stack is standard: accessibility-snapshot observations with stable references, and fixed tool calling over primitives plus skill invocations. Compared with the state-of-the-art approaches, SkillMigrator reduces the average LLM-action count on successful trajectories by 8-10% across both WebArena and Mind2Web at matched success rate.

Generalist embodied agents require more than object recognition: they must reason about spatial relations, actions, procedures, human intentions, environmental constraints, and commonsense consequences from situated visual observations. Yet existing visual and embodied question answering benchmarks often provide limited control over the reasoning dependencies being tested, making it difficult to distinguish grounded embodied reasoning from shortcut-driven visual or linguistic pattern matching. We present ERQA-Plus, a diagnostic benchmark for reasoning in embodied AI. ERQA-Plus contains 1,766 question-answer instances grounded in 711 robot-centric images and organized according to a structured taxonomy spanning perceptual, action-centric, social-interaction, navigation-environmental, and contextual commonsense reasoning. The dataset is constructed using a multi-stage generation and validation pipeline that combines taxonomy-guided question generation, automatic quality judging, iterative revision, and human assessment to improve visual grounding, answer validity, and reasoning quality. We benchmark representative general-purpose vision-language models and embodied models, including LLaVA-NeXT-8B, Prismatic-7B, MiniCPM-V-4.5-8B, Qwen3-VL, RoboRefer-8B, and RoboBrain2.5-8B. Although the strongest model, Qwen3-VL-32B, achieves 83.4% overall accuracy and 61.4 SBERT score, category-level results reveal persistent weaknesses in spatial reasoning, procedural reasoning, event prediction, and intention inference. ERQA-Plus therefore provides a fine-grained evaluation framework for measuring not only whether embodied agents answer correctly, but also which forms of embodied reasoning they can and cannot perform reliably. The dataset is available https://huggingface.co/datasets/huggingdas/erqa-plus and the project page at https://github.com/LUNAProject22/erqa-plus.

Evaluating large language models (LLMs) is important for understanding their capabilities, comparing competing systems, and supporting the deployment of reliable models in practice. For open-ended tasks, pairwise evaluation has become a popular paradigm, in which two responses to the same prompt are compared and the resulting judgments are aggregated into an overall ranking. A central challenge of this paradigm is intransitivity: the induced comparison outcomes may fail to support any coherent global ranking. For example, one may observe cyclic preferences such as $A \succ B \succ C \succ A$, or inconsistencies involving ties such as $A \equiv B\equiv C\neq A$. Such contradictions make the resulting leaderboard unstable and challenging to interpret. In this paper, we propose a prompt perturbation framework for improving the consistency of pairwise LLM evaluation. Our approach generates perturbed variants of each prompt, uses the resulting comparison graphs to identify and filter out structurally inconsistent comparison patterns, and then applies standard ranking methods to the filtered comparisons. A key feature of the proposed framework is that graph-level structural consistency is incorporated explicitly into the evaluation pipeline before ranking aggregation. This provides a simple and principled way to reduce cyclic inconsistencies and improve the reliability of LLM rankings.

Memory has become a standard substrate for self-evolving agents, yet retaining experience is not the same as learning how to evolve through it. Existing memory agents can store trajectories, retrieve reflections, or accumulate skills, but often lack the holistic competence to select useful experience, act on it, write reusable knowledge, and maintain a growing repository. We introduce OPD-Evolver, a slow-fast co-evolution framework that cultivates such an agent evolver through on-policy self-distillation. In the fast loop, OPD-Evolver interacts with a four-level memory hierarchy to read, use, write, and maintain experience for rapid test-time evolution. In the slow loop, outcome-calibrated memory attribution and privileged hindsight distill these four abilities into the deployable policy. Across multi-domain benchmarks, OPD-Evolver surpasses memory systems such as ReasoningBank by up to 11.5%, and training-based methods such as Skill0 by ~5.8%. Further analysis shows that OPD-Evolver internalizes high-value experience and memory management, enabling OPD-Evolver-9B to challenge giant counterparts such as Qwen3.5-397B-A17B and Step-3.5-Flash, pointing beyond memory-augmented agents toward genuinely qualified agent evolvers.

Fine-grained action recognition in egocentric video is challenging for Vision-Language Models (VLMs): actions often differ only in small visual cues, and a single model tends to be biased toward a subset of these cues. We propose Divide, Deliberate, Decide, a fully-local, zero-shot multi-agent framework in which (i) a VLM orchestrator chunks the video and proposes a top-k candidate label list per segment, (ii) an ensemble of heterogeneous VLM specialists, drawn from different open model families, engages in a structured deliberation that includes a peer-consultation round of questions, and (iii) agent rankings are aggregated with a Borda count and the orchestrator re-ranks its own prediction in light of the specialists' evidence. The entire pipeline runs locally with no fine-tuning. Experiments show that our method positively improves zero-shot action recognition performance over the baseline, highlighting the influence of a heterogeneous deliberation step, showing that the gain stems from decorrelated model priors rather than from additional compute.

Reference-driven image generation has made rapid progress on identity preservation, but reliable viewpoint control across different subjects remains poorly understood. The difficulty is not merely generating a new image of the target subject: the model must infer the implicit viewpoint of one subject and transfer it to another subject using only image-level evidence, without camera poses, depth, or ray-based conditions. In this setting, existing generators conditioned on multiple image references often rely on spurious semantic correlations, which lead to viewpoint drift, part-level structural mismatches, and missing or unsupported target-specific content. We formulate this challenge as cross-subject viewpoint alignment and propose RAVA, a retrieval-augmented framework that supplies explicit geometric evidence before generation. RAVA first learns a cross-instance viewpoint embedding that retrieves target-subject images aligned with the anchor viewpoint, then applies a LogDet-based subset selection strategy to retain a compact reference set that is both view-consistent and structurally complementary. The selected references are finally consumed by a fine-tuned multi-reference image generator. Experiments show that generic semantic embeddings are nearly random for this task, while the proposed retriever substantially improves viewpoint retrieval quality. On cross-subject generation, RAVA consistently outperforms zero-shot baselines and stronger retrieval alternatives under the same generation backbone. These results indicate that cross-subject viewpoint alignment benefits from retrieval-augmented geometric grounding rather than relying on end-to-end generation alone.

Compressing large language models reduces memory use and inference cost, but it can also create failures that standard benchmarks miss. A pruned model may still perform well on multiple-choice evaluations, yet fail to answer the same question in open generation. We ask what pruning changes: does it erase the correct answer, or does it make the answer harder to produce as the top output? We study this question with multilingual question answering, tracking the same questions before and after pruning. We find a benchmark illusion. Under high-sparsity pruning, especially Wanda, models often fail in greedy open generation while still selecting the correct answer under multiple-choice scoring. In these recognition-only errors, the answer is usually not gone, but demoted: it often reappears with beam search, sampling, or one in-context example. Overall, multiple-choice benchmarks can overstate the usability of compressed LLMs, creating an evaluation blind spot. Compressed models should be tested on what they can produce, not only on what they can recognize.

The widespread deployment of face recognition (FR) systems exposes personal images shared on social media and public platforms to identity linkage and privacy risks. Existing adversarial privacy protection methods can degrade unauthorized FR performance but are not compatible with generative face editing. Artificial intelligence-driven face editing tools are gaining popularity, which has significantly increased user demand for personalized portrait generation and social sharing. However, current editing methods often preserve identity features, making the edited images still susceptible to tracking by malicious FR systems. Thus, this paper proposes Flux-Guard, a privacy-preserving face editing framework based on adversarial attacks, which integrates face editing and privacy protection within a unified generative process. Specifically, we design a flow trajectory control method to align semantic manipulations with the generative process and introduce latent-space adversarial optimization with an adaptive perceptual-loss-driven weighting strategy, dynamically adjusting adversarial strength to maximize attack effectiveness while preserving visual quality. Extensive experiments demonstrate that Flux-Guard supports face editing while significantly improving attack success rates against cross-domain face recognition models on the CelebA-HQ and LADN datasets. Furthermore, evaluation results for commercial APIs have confirmed its effectiveness in real-world applications. The code is released at https://github.com/JLMWang/Flux-Guard.

Text-guided Open-vocabulary Object Counting (TOOC) enables counting arbitrary object categories specified by text prompts, offering substantially greater flexibility than conventional closed-set counting. However, existing TOOC methods are developed and evaluated primarily on ideal images, while real-world scenes often suffer from adverse conditions such as rain, fog, darkness, and sensor noise, which severely degrade visual quality and impair vision-language alignment. To bridge this gap, we introduce Robust-TOOC, the first benchmark for evaluating TOOC under diverse corruption conditions, which covers six representative degradation types: rain, fog, darkness, Gaussian noise, salt-and-pepper noise, and mixed corruption. To improve robustness while preserving the original counting architecture, we propose Dual-TTT, a dual-architecture test-time training framework for TOOC. Specifically, during test-time training, Dual-TTT updates only the Text-guided Lightweight Denoising module (TL-Denoiser), while keeping the original counting network frozen. Inspired by diffusion models, the TL-Denoiser is optimized to remove corruption-aware noise from image representations under degraded conditions. Since only the TL-Denoiser is trained at test time, Dual-TTT is annotation-free and can be seamlessly integrated into existing TOOC models without modifying their original architecture. Extensive experiments on multiple recent TOOC baselines demonstrate the effectiveness of our method.

Humans naturally leverage diverse sensing modalities to interact with the physical world, while most Vision-Language-Action (VLA) models for robotics rely solely on RGB observations. This limits their ability to perceive physical properties that are difficult or impossible to infer from RGB cameras, such as temperature, sound, or radar response. We present MuseVLA, an adaptive multimodal sensing VLA model that integrates novel sensors as on-demand tools for robotic manipulation. Given a task instruction and visual context, MuseVLA first generates a sensor token and target description that select the sensing modality to invoke and what to attend to, analogous to a tool call with arguments. It then converts the selected sensor measurement into a grounded sensor image, a unified intermediate representation that encodes heterogeneous readings for multimodal fusion and action generation. This design decouples sensor-specific processing from the VLA backbone, enabling efficient integration of diverse modalities. To reduce the need for expensive multisensory robot datasets, we further introduce a data synthesis pipeline that augments existing RGB video datasets with grounded sensor images, enabling generalization to unseen sensor-guided tasks. We evaluate MuseVLA on a real-world robot across challenging dexterous hand manipulation tasks that require multimodal sensing inputs, including temperature-guided pick-and-place, audio-driven object search, and radar-assisted hidden object retrieval. MuseVLA achieves 80.6% success rate on average, outperforming RGB-only and multisensory VLA baselines significantly, and exhibits strong zero-shot capabilities on unseen tasks.

Video tokenization is fundamental to scalable video generation, as the number of tokens directly determines the computational cost and the length of videos that can be modeled. Existing tokenizers mainly improve scalability by compressing videos into fewer tokens, but they often continue to represent persistent content, such as static backgrounds and consistent object appearances, repeatedly across frames and chunks. In this paper, we propose \textbf{TivTok} (\textit{Time-Invariant Tokenizer}), a reuse-aware video tokenizer that makes persistent information reusable across time. TivTok represents a clip with Time-Invariant (TIV) tokens that encode information shared across frames and Time-Variant (TV) tokens that encode frame-specific residuals. To obtain this factorization, we introduce Scope-Induced Factorization (SIF), which assigns different attention scopes to the two token groups: TIV tokens attend to the full clip, whereas each TV token only accesses its corresponding frame together with the TIV tokens. In the decoder, Invariant Broadcasting (IB) reuses the same TIV tokens across frames and chunks for parallel reconstruction and long-video tokenization. Experiments show that TivTok achieves an rFVD of 12.65 on the standard $16{\times}256{\times}256$ benchmark and improves compression efficiency by 2.91$\times$ for 128-frame videos compared with the evaluated baselines, while using only 1.1\% of the tokens required by downsample-based tokenizers in our evaluation.

Adding LLM-generated node features to graph neural networks (GNNs) is widely reported to improve accuracy on standard benchmarks. We document a contrasting observation: when LLM features are introduced through pure input concatenation (rather than joint training, distillation, or prompt-conditioning), they can systematically degrade accuracy on the same homophilous benchmarks where end-to-end LLM pipelines succeed. With an MLP backbone on the Planetoid public split and bag-of-words original features, concatenating SBERT-encoded GPT-4o-mini TAPE features reduces PubMed test accuracy by -17.0 +/- 0.3 pp and Cora by -4.3 +/- 0.6 pp (CiteSeer -0.6 +/- 0.8 pp, within seed noise). The drop attenuates as we relax each condition (GCN / GCNII / GAT backbones, random splits, smaller encoders) and reverses on medium-homophily WikiCS (+4.4 pp) and ogbn-arxiv (+11.7 pp). To predict when concatenation helps versus hurts, we report a simple measure of LLM-alone discriminability, Delta_sig. Across 9 datasets Delta_sig correlates with the concatenation cost more strongly than homophily at point estimate (r^2 = 0.38 vs. 0.06; N=9, bootstrap CIs overlap). The bootstrap-best change-point is tau = 13.8 pp, and the rule "Delta_sig <= tau predicts non-positive concat cost" classifies 7/9 datasets correctly; since 60% of bootstrap samples place tau in [5, 30] pp, we treat Delta_sig as an interpretive lens rather than a precision filter. A dimension-controlled ablation on PubMed places the LLM-feature drop between same-source PCA (-2.3 pp) and same-dim Gaussian noise (-37.3 pp), ruling out dimensionality and weight-decay artifacts. Nine PubMed configurations fit a power law |Delta_concat| proportional to (sqrt(d_l/n))^1.31 with r^2 = 0.97; the low-Delta_sig, small-n corner is exactly where the headline -17 pp PubMed deficit appears.

While sequential residual fitting is the bedrock of standard boosting frameworks, it inherently breeds learner redundancy by repeatedly revisiting correlated error components. To address this bottleneck, we propose a shift from residual fitting to \textit{residual orthogonalization} and introduce SCBoost. Our framework tackles redundancy through two complementary mechanisms: Spectral Residual Projection (SRP) and Covariance-Regularized Weighting (CRW). During training, SRP projects each residual target onto the orthogonal complement of the historical prediction subspace, forcing successive learners to capture only novel empirical innovations. During aggregation, CRW optimizes ensemble weights on a validation set with an explicit covariance penalty to mitigate remaining correlations. Theoretically, we provide a finite-sample geometric characterization proving that SRP yields an exact additive residual-energy decomposition. Furthermore, under an isotropic-noise assumption, we rigorously establish the conditions under which this projection improves the effective Signal-to-Noise Ratio. Extensive experiments across ten benchmark datasets demonstrate that SCBoost delivers strong out-of-the-box performance, particularly in accuracy and F1 score. This work reinterprets boosting through a geometric lens, suggesting that explicit redundancy control is a principled and necessary step toward more efficient ensemble architectures.

Standardized evaluation protocols are indispensable for robust benchmarking in remote sensing, particularly as foundation features are increasingly transferred across diverse sensors and complex imaging geometries. In satellite multi-view reconstruction, conventional evaluations relying on unconstrained 2D global matching are often misleading. The Rational Function Model (RFM) and its Rational Polynomial Coefficients (RPC) dictate a curved, height-dependent epipolar geometry that render flat 2D search spaces physically inconsistent. We propose a geometry-faithful and reproducible protocol tailored for the RPC framework. Our approach integrates an RPC-projected 3D consistency metric with a geometry-constrained dense matching proxy, specifically evaluating whether similarity responses remain localized and unique under physically plausible search manifolds. A pivotal finding of our joint reporting strategy is the decoupling of semantic agreement and geometric localization: high cross-view similarity at a projected 3D point does not guarantee reliable matchability in practical inference. Our benchmark demonstrates that incorporating geometric constraints is fundamental to the problem definition in satellite imagery. Furthermore, we show that state-of-the-art 2D backbones remain remarkably competitive against specialized 3D-aware models when subjected to this RPC-consistent evaluation.

LiDAR sensors are widely deployed in autonomous systems for 3D perception and safety-critical decision-making. We identify a previously unexplored attack surface in which dormant malware embedded in the LiDAR sensing pipeline remains inactive during normal operation and can be externally triggered after deployment, without requiring access to sensor hardware or networking at attack time. To operationalize this threat, we design malware capable of low-level point-cloud manipulation and embed it into LiDAR firmware. This malware was developed in a closed research test environment with vendor technical support, rather than by exploiting an inherent production supply-chain vulnerability. To selectively trigger attack activation, we design and implement an optical trigger that remotely activates the malware by delivering a modulated signal into the sensing environment. Once triggered, the malware performs real-time point cloud manipulation, and we demonstrate false object injection and real object suppression on static and mobile victim platforms. Our evaluation first establishes attack feasibility, including static operation at 300~ft and recorded drive-by runs reaching 35~mph. We then illustrate quantitatively that injected person-like artifacts can remain semantically detectable by a state-of-the-art 3D object detector. Finally, we demonstrate multiple modes of safety-critical impact on a deployed tactical autonomous vehicle. Together, these results highlight the need for stronger integrity guarantees throughout the LiDAR sensor development and deployment pipeline.

Image restoration seeks to recover high-quality images from degraded inputs but becomes highly ill-posed under complex, mixed degradations. While unified all-in-one models are common, their performance declines as degradation complexity increases. Recent works adopt Chain-of-Thought (CoT) reasoning for multi-round restoration using specialized modules. However, this approach faces two key limitations: (i) increased computational cost due to multi-step processing, and (ii) weak modeling of interactions between degradations during stepwise inference. We introduce CoTIR, a universal image restoration framework that internalizes CoT reasoning within a single model. Concretely, we view image restoration as a specialized subtask of image editing, which implies that a large-scale pre-trained editing model provides a more favorable optimization starting point. Building on this, we fine-tune the model for restoration and further encode structured CoT-style reasoning into the learning objective via a differentiable formulation inspired by Lagrangian optimization, enabling holistic restoration without chaining specialized restorers. To facilitate training and evaluation, we further present CoTIR-Bench, a large-scale benchmark comprising 5.2 million samples with CoT-style reasoning traces. Extensive experiments on CoTIR-Bench and broad real composite degradation scenes show that CoTIR achieves stronger perceptual quality and more competitive fidelity than both all-in-one models and multi-round restoration methods. The source code is available at https://github.com/gy65896/CoTIR.

Geographic tipping points in ecosystems, climate subsystems, or ice sheets pose severe challenges for localized early warning. Classical spatial indicators such as Moran's I summarize global spatial structure, but they struggle with three issues: spatial dilution, Euclidean assumptions, and correlated noise. This paper introduces SpatioTemporal Causal Network Diagnostics (ST-CND), a framework that addresses these three issues by representing the geographic field as a time-evolving directed causal network. The core workflow is as follows: (1) infer which spatial nodes help predict other nodes via transfer entropy, replacing fixed Euclidean neighborhoods with data-driven information-flow topology; (2) estimate local recovery rates within each candidate subnetwork via dynamic mode decomposition; and (3) identify the most vulnerable subnetwork by combining three signals, namely high internal fluctuation, high internal synchronization, and low external coupling, thereby suppressing false alarms from spatially correlated noise. Validated on synthetic bifurcations and two observational sea-surface temperature benchmarks, namely Indo-Pacific SST and North Atlantic AMOC, ST-CND delivers localized and interpretable warnings. On the AMOC task, it achieves an AUROC of 0.783 and a critical-subnetwork IoU of 0.378, outperforming recurrence-network and lambda-AR1 baselines. The framework provides an interpretable and scalable pipeline for spatial early warning in Earth system science.

Simulation based solvers for optimal stopping problems must discretize the stopping decision. Under classical dynamic programming, a coarse exercise grid with only a few stopping opportunities can materially undervalue the optimal expected reward, whereas on a very fine grid, approximation errors accumulate through the backward recursion. To remove this limitation, we develop a new reinforcement-learning inspired algorithm that enables us to learn the exercise rule at arbitrarily fine time resolution. Our CARLOS (Continuous-time Adaptive Reinforcement Learning for Optimal Stopping) algorithm utilizes an aggregate deep neural network (ADNN) to learn a joint space-time decision boundary. Starting from a coarse time grid, we progressively increase the frequency of stopping opportunities, while in parallel training the ADNN to refine its timing-value estimates. We moreover design an adaptive sampling strategy that gradually concentrates training effort near the stopping boundary. Benchmarked results show that CARLOS delivers higher prices than existing Bermudan solvers, approaching the American upper bound, and achieves high computational efficiency relative to non-RL comparators.

We investigate turn-taking in multimodal multi-party conversations using large language models (LLMs). We construct an evaluation framework for three tasks: addressee detection, turn-change prediction, and next speaker prediction. We compare supervised models trained for these tasks, text-based LLMs, multimodal LLMs (MM-LLMs), and human subjects. Experiments on the AMI corpus showed that LLMs outperformed supervised models and humans in next speaker prediction, despite not being trained on the target domain and without access to audio or visual information. An MM-LLM performed better than text-based LLMs on addressee detection and turn-change prediction but remained below human performance, indicating difficulty leveraging raw audio-visual signals. Ablation analyses revealed that conversational context was critical, particularly for next speaker prediction. We observed that human and LLM prediction patterns were similar, and intervals with frequent turn changes were difficult for both.

Offline evaluation of agentic systems often collapses trajectories to terminal success, discarding information about partial progress and inducing widespread ties, creating substantial statistical inefficiency by reducing effective sample size and weakening the ability to distinguish systems. We propose preference-based trajectory evaluation, which compares trajectories directly through temporal preferences over progress and time-to-return profiles. We find that, across diverse agentic and interactive benchmarks, standard success-based metrics produce tied comparisons on roughly 75% of instances, whereas trajectory-aware preferences reduce ties to roughly 35%, improving discriminative power, ranking stability, and data efficiency. Our results suggest that benchmark saturation, often attributed to poor data collection or problem difficulty, may also be explained by the choice of evaluation measure.

Multi-threat robustness remains a fundamental challenge in deep learning. Although joint adversarial training (JAT) is widely adopted, it suffers from negative transfer under heterogeneous threats, particularly between $\ell_p$-bounded and semantic attacks. Through first-order gradient analysis, we formalize this as gradient incompatibility and theoretically establish the necessity of decoupled optimization. We further reveal that these conflicting threats exhibit separable spectral characteristics in the frequency domain. Motivated by this observation, we propose Threat-aware Frequency Decoupling (TaFD), a two-stage defense framework that reformulates JAT as a frequency-domain divide-and-conquer paradigm. TaFD first discovers latent threat domains via unsupervised clustering of attack spectral prototypes and trains a lightweight classifier for inference-time threat domain identification. Conditioned on the prediction, TaFD employs a Frequency-Conditional Convolution that learns threat-domain-specific spectral masks and routes each sample to the corresponding expert, enforcing structural parameter separation and alleviating optimization conflicts. We validate TaFD on three representative image-classification benchmarks (CIFAR-10, CIFAR-100, and Tiny-ImageNet) and on two representative architectures (the convolutional ResNet and the hybrid-transformer MobileViT). Extensive results demonstrate that TaFD achieves more balanced robustness against heterogeneous attacks than existing JAT and frequency-domain baselines, improving average robust accuracy by approximately 11\% over the strongest baseline while maintaining leading clean accuracy.

Spatial VLMs have made substantial progress in geometric perception, yet complex spatial reasoning requiring multi-step inference over depth, distance, and scene relations remains challenging. Moreover, different spatial queries call for fundamentally different strategies: some are best addressed through purely linguistic, step-by-step deduction, while others require explicit 3D grounding before quantitative inference. We present Dual-Path Spatial Reasoning via Reinforcement Learning for Spatial VLMs (SR-REAL), a unified framework that equips a spatial VLM with two complementary reasoning paths: Language-Only Reasoning (LOR), which performs step-by-step linguistic deduction, and Detect-Then-Reason (DTR), which detects 3D geometric cues (e.g., centers or bounding boxes) via region tokens before explicit geometric inference. SR-REAL begins with a cold-start supervised fine-tuning stage that constructs LOR and DTR chain-of-thought supervision and exposes a region-to-3D interface, followed by RL that optimizes the policy model with accuracy and format rewards; for DTR, a discrete center-based detection reward further refines geometric alignment. Across diverse spatial benchmarks, SR-REAL significantly outperforms spatial VLM baselines: (i) a single RL-trained model supports both reasoning paths, with DTR excelling in region-aware tasks through precise 3D localization and LOR enhancing general spatial reasoning; (ii) jointly training both paths fosters mutual reinforcement; (iii) high-quality, blended cold-start data is crucial for stable RL optimization; and (iv) the model generalizes across datasets and domains without per-task tuning, demonstrating positive transfer between LOR and DTR.

Generative world models for autonomous driving face two unresolved tensions: heterogeneous control injection, where free-form language, HD-maps, trajectories, and camera poses reside in incompatible representational spaces, and post-hoc cross-view fusion, where per-camera latents fail to encode global 3-D geometry. We trace both to a single root cause: the absence of a shared symbolic interlingua aligning language, geometry, and pixels at the latent-token level. We present DRIVE-CHOREO, an LLM-choreographed multi-agent world model that recasts controllable multi-view video generation as latent choreography. Three Qwen2.5-VL agents - a Director parsing user intent into a structured WorldScript, a Cartographer grounding it into spatially-anchored layout tokens, and an Auditor feeding cross-view critiques back as auxiliary supervision - jointly author a single position-aware token sequence. This sequence is co-compressed with the multi-view video via a view-time permutation that enforces inter-camera geometry within the convolutional receptive field of a 3-D VAE. On nuScenes, DRIVE-CHOREO sets new state-of-the-art multi-view consistency and BEV mAP (21.6) with competitive FVD (45.7); a detector trained purely on our synthetic data gains +2.4 NDS on the real validation split, validating downstream utility.

Large language models show strong reasoning ability, but their internal reasoning process can remain unstable in complex multi-step settings, where early hidden-state errors may propagate to incorrect predictions. We propose ReLAR, a reinforcement-guided latent refinement framework that iteratively updates hidden representations before decoding. ReLAR maintains a compact latent reasoning state and uses learned depth and action controllers to adaptively determine both the number and direction of refinement steps. The controllers are trained with a policy gradient objective based on step-wise likelihood improvement, enabling efficient input-dependent reasoning without explicit chain-of-thought generation. Experiments on medical, mathematical, multi-hop reasoning, and open-ended generation benchmarks show that ReLAR improves accuracy, generation quality, and reasoning stability with substantially lower inference overhead than explicit reasoning baselines.

Training embodied agents in the real world requires skilled operators and expensive hardware. Simulation environments offer a compelling alternative by enabling large-scale, cost-effective data augmentation. Consequently, rapidly constructing high-fidelity simulation scenes with a minimal sim-to-real gap has become a critical objective in robot learning. While reconstruction-based methods provide superior visual quality, current workflows are hindered by inefficient data acquisition and subpar foreground object extraction. We thus propose GASE, a highly automated system for simulation scene construction. GASE leverages multi-view video streams from panoramic camera arrays to enable rapid environment scanning. To ensure high-quality asset generation, our pipeline introduces a camera-pose-based strategy that robustly extracts objects across frames in the 2D domain, followed by high-fidelity scene inpainting. Foreground objects and the static background are then reconstructed independently and seamlessly imported into physics simulators for policy training. Extensive experiments demonstrate that GASE outperforms existing 3D Gaussian-based methods in segmentation accuracy by over 10\% while achieving state-of-the-art inpainting quality. Furthermore, real-robot deployments across manipulation and navigation tasks maintains a performance gap of less than 10\% compared to policies trained purely on real-world data. These results confirm that GASE provides an efficient and highly effective solution for bridging the sim-to-real gap. Code will be released.

Production LLM assistants route user requests to growing libraries of specialized tools, but how does routing accuracy degrade as the catalog scales? We study single-step routing on a 110-agent, 584-tool catalog from a deployed enterprise productivity assistant, evaluating three frontier models from 10 to 110 agents. Routing F1 on under-specified requests drops 16--23 percentage points across models. An oracle analysis decomposes the degradation into a \emph{retrieval} gap (the model cannot surface the right tool) and a \emph{confusion} gap (even with perfect retrieval, the oracle ceiling drops 10pp). Embedding-based shortlisting recovers +10--11pp F1 at full scale across all three models and two providers. A production annotation study (1,435 human-labeled utterances, three annotators) confirms the recovery on real traffic at +10--17pp despite 10--15pp lower absolute performance.

Neural operators provide fast surrogates for PDEs but their deterministic predictions limit their use in tasks requiring uncertainty quantification (UQ), especially under geometric variability. Existing approaches primarily model uncertainty in network parameters, largely overlooking the geometry-aware representations learned by the operator itself. We propose REEF-GP (Residual on Embedded Features Gaussian Process), a post-hoc UQ framework that fits a GP to the residuals of a frozen neural operator whose internal embeddings define the kernel feature space. Rather than learning a separate feature map, REEF-GP adapts the operator's intrinsic coordinate-feature representations to construct geometry-aware uncertainties. To ensure stability and scalability on unstructured domains, REEF-GP incorporates spectral-normalized projections, heteroscedastic geometry-aware noise, and efficient subset-based training that avoids restrictive low-rank approximations. Across five PDE benchmarks with varying geometries, REEF-GP preserves predictive accuracy while achieving calibrated uncertainty estimates competitive with deep ensembles but at a fraction of their cost. Our approach remains robust under geometric distribution shift, with uncertainty concentrating in physically meaningful regions (e.g., shock fronts). Our results demonstrate that accurate and scalable post-hoc UQ for neural operators can be achieved directly in their learned feature space, offering a practical alternative to parameter-centric approaches.

Robot learning and embodied agents now require simulation to serve as a shared execution substrate linking control, skills, and planning, not only as a renderer, controller testbed, or fixed task environment. Existing pipelines split these layers with "magic" actions, disconnected training environments, or forward-only renders that cannot reproduce, evaluate, and annotate the same episode. We present MagicSim, an embodied interaction infrastructure built around one deterministic batched runtime and a shared Markov decision process (MDP). From YAML-first specifications that decouple contents, placement, behavior, and agent exposure, MagicSim constructs diverse executable worlds spanning task families, interaction regimes, physics, layouts, sensors, avatars, and robot embodiments in one reset-and-step loop. A common execution interface grounds high-level commands through controllers, atomicskills, planner primitives, and asynchronous planning, realizing them as robot actions rather than simulator-side state edits. One task definition supports three capabilities: benchmark and RL evaluation, an autocollect interface that automatically turns commands into grounded trajectories, and agent/VLM-facing interaction. For automatic execution, commands flow through a Command->Skill->Planner->Robot->Record pipeline, while per-environment command, skill, planning, retry, annotation, and episode states advance independently above the shared physics tick. Successful rollouts are saved as structured multimodal trajectories aligning language supervision, action representations, visual/geometric representations, and task-level status with the executed episode. MagicSim thus unifies diverse world construction, embodied execution, task evaluation, automatic rollout generation, and interactive agent interfaces in one planner-in-the-loop runtime.

Despite rapid advances in UAV technologies, current deployments remain limited due to several gaps in UAV systems research. To address these challenges, we propose OmniDroneX, a unified Drone-as-a-Service ecosystem, in which drones are transitioned from fixed function platforms into dynamically composable entities that can be integrated with external infrastructures to offer omni-capabilities. OmniDroneX bridges low-level physical primitives with high-level mission intent through a unified vendor-agnostic interface (libUAV) and a formal physical-service abstraction model (PT-SOA). A core innovation is the diverse application of large language models (LLMs) across multiple layers of the OmniDroneX architecture. LLMs are used to assist in identifying and formalizing primitive device functions and abstract service definitions, supporting automated service composition and workflow generation, and enabling interactive, natural-language mission specification and refinement. OmniDroneX also incorporates important categories of composition techniques that are essential in dynamic UAV systems, including physical layer composition for drone capability augmentation, as well as spatiotemporal, functional, collaborative, exception-aware, and QoS-based service compositions. Collectively, these features allow OmniDroneX to serve as a foundation for scalable, resilient, and self-evolving UAV ecosystems operating in complex and dynamic environments.

Large Language Models (LLMs) are increasingly deployed in edge-cloud inference systems to handle diverse user tasks with heterogeneous accuracy, latency, and cost profiles. Selecting the appropriate LLM for each incoming task is critical for ensuring service quality and efficient resource utilization. However, model heterogeneity, stochastic and unknown performance characteristics, and time-varying task demands make static selection strategies inadequate. Real-world deployments often impose hard resource budgets such as monetary expenditure limits, along with soft service-level requirements such as latency guarantees. These constraints introduce additional challenges for online decision-making. We formulate this problem as a constrained stochastic bandit learning task, where the learner sequentially selects models under both packing-type (hard) and covering-type (soft) constraints, while adapting to time-varying task demand. The learner operates without access to the underlying reward, cost, or latency distributions and must rely on partial feedback. We develop a novel online learning algorithm that leverages confidence-bound estimates and demand predictions to balance reward maximization with long-term constraint satisfaction. We provide theoretical guarantees showing sublinear regret and sublinear covering constraint violations compared to an offline benchmark with full information. Experimental results on synthetic workloads demonstrate the effectiveness and robustness of our approach in dynamic, resource-constrained environments.

Generating adversarial driving scenarios is critical for evaluating and improving autonomous vehicle decision-making systems in simulation. Recent approaches, such as ChatScene and LLM-Attacker, rely primarily on the prior knowledge of Large Language Models and Vision-Language Models to generate driving scenarios procedurally. We argue that adversarial scenes should be generated based on the failure diagnosis (e.g., indecisiveness, multi-frame inconsistency) of the driving policy to specifically address the policy's weaknesses instead of relying on prior assumptions. In this paper, we propose SPHINX, a closed-loop framework for adversarial scenario synthesis guided by a simple principle: first explain, then explore. Beyond blindly exploring the scenario space, SPHINX leverages explainable artificial intelligence methods to analyze the policy, identifying key visual concepts and their influence on policy outputs, and the uncertainty of the decisions. Given the interpretable evidence extracted from the policy's own decision process, we use a vision language model to rationalize and criticize failure modes of the current policy. These critics are then used to generate targeted adversarial scenarios for policy retraining and improvement. We demonstrate that SPHINX can highlight an interpretable account of policy failures while other adversarial scene generation cannot. Across the evaluated benchmarks and test suites, SPHINX can be applied to diverse state-of-the-art autonomous vehicle architectures and yields consistent robustness improvements over existing scenario-generation methods.

Generalist vision-language-action systems need object-centric 3D evidence and reusable manipulation experience to plan reliable robot trajectories. GeneralVLA provides a hierarchical interface for converting language and RGB-D observations into 3D end-effector paths, but two bottlenecks remain. First, monocular SAM3D-style object reconstruction can hallucinate pose and unseen geometry, while manipulation benefits from stable object shape when calibrated multi-view observations are available. Second, the original KnowledgeBank mainly retrieves semantically similar snippets and appends new knowledge, which makes it difficult to control memory quality, conflicts, confidence, and geometric relevance. To address the first challenge, we introduce GeoFuse-MV3D, a geometry-prior-guided MV-SAM3D reconstruction branch that verifies external geometry cues with input-view masks, applies soft visual-hull support, performs axis-wise refinement, and fuses only geometry while preserving appearance. To address the second challenge, we upgrade KnowledgeBank into a governed long-term memory system with explicit quality, confidence, lifecycle, verifier, and conflict metadata, together with precision-oriented retrieval. Finally, we evaluate the reconstruction branch on GSO-30 and the memory module on Terminal-Bench 2.0 and SWE-Bench Verified; GeoFuse-MV3D improves over the MV-SAM3D baseline by reducing CD and LPIPS by 2.20% and 2.02% while increasing PSNR and SSIM by 2.36% and 1.03%, and KnowledgeBank improves over ReasoningBank by 4.53% on Terminal-Bench SR and 3.73% on SWE-Bench resolve rate, while reducing AS by 4.95% and 5.65%, respectively. Code: https://github.com/AIGeeksGroup/GeneralVLA-2. Website: https://aigeeksgroup.github.io/GeneralVLA-2.

As LLMs acquire stronger reasoning capabilities, deceptive behavior becomes an increasingly serious safety concern. Existing deception monitors either score visible transcripts or derive scalar probe scores from representation vectors, leaving little inspectable evidence about why a response is suspicious. We introduce STATEWITNESS, an activation explainer for deception auditing. A separate decoder reads a target model's hidden states, then answers natural-language queries or emits structured reports about them. We evaluate STATEWITNESS on two target reasoning LLMs across seven deception datasets. STATEWITNESS reaches 0.916 mean AUROC, a relative gain of 11.6% over the best black-box text monitor and 25.0% over the best activation-probe baseline under the same evaluation protocol. When combined with existing monitors, STATEWITNESS reduces missed deceptive examples in simple threshold ensembles. Beyond scalar detection, the decoder returns query-level answers, schema reports, and token- or sentence-level evidence traces for human inspection. We view this interface as a potential building block for broader interpretability and alignment tools.

Stereo matching has advanced through foundation models trained on large-scale datasets, yet this paradigm suffers from a scalability bottleneck: incorporating new data requires costly joint retraining. Model merging offers a scalable post-hoc alternative by integrating knowledge from specialized models after source checkpoints are available. However, existing merging methods typically retain all available models or rely on greedy inclusion, which can preserve harmful task-vector interference. We propose StereoFactory, a coarse-to-fine evolutionary framework for adaptive model merging. Stage~1 employs a genetic algorithm to search the combinatorial space of model subsets, determining which models should participate. Stage~2 addresses module-level knowledge specialization (different functional modules exhibit distinct preferences for knowledge sources) through CMA-ES optimization of architecture-adaptive routing over the selected task vectors, with optional module-level scaling. Experiments across two architectures and four benchmarks demonstrate that StereoFactory consistently achieves the best four-benchmark average under the same checkpoint pool, reducing the average error from 3.80 to 3.30 on NMRF and from 2.88 to 2.19 on FoundationStereo relative to the strongest controlled baseline. The post-hoc search requires only 2.7--3.7\% of the corresponding joint-retraining wall-clock time. Analysis reveals that knowledge contributions are inherently module-specific, and selected subsets can transfer across architectures with minimal degradation. Code will be publicly released upon acceptance at: https://github.com/XiandaGuo/StereoFactory.

Large language models (LLMs) are increasingly considered for use in clinical consultation tasks, yet most medical evaluations remain static, single-turn, or narrowly outcome-based, limiting their ability to reflect the sequential, uncertain, and interactive nature of real-world care. Here, we propose AIPatient Arena, an EHRs-grounded evaluation framework for assessing the clinical utility of LLMs across eight dimensions of clinical competence. The framework integrates EHR data into patient-specific knowledge graphs, enabling multi-turn physician-patient interactions. We applied AIPatient Arena on a primary cohort of 437 patients and two out-of-distribution validation cohorts of 119 and 67 patients. We observe that LLMs performed well in medical interview questioning skills (QS; mean scores, 4.43-4.99/5), ethical and professional conduct (ET; 4.38-4.93/5), and clarity and transparency of clinical explanations (EX; 3.80-4.72/5). Performance was moderate in information integration (II; 3.19-4.21/5) and medication safety and justification (MS; 3.13-3.78/5), but persistent weaknesses were observed in handling of ambiguous patient responses (HR; 2.57-3.32/5), information coverage (IC; 2.08-3.02/5), and diagnostic accuracy and reasoning (Dx; 2.63-3.55/5). Process-based evaluation revealed recurrent interaction failures, including repetitive questioning, omission of past medical history, and inadequate handling of uncertainty. Richer conversational context improved diagnostic reasoning but yielded limited gains in treatment planning. These findings indicate that final-answer accuracy alone is insufficient for evaluating clinical readiness and highlight the importance of assessing how models gather, interpret, and communicate information throughout a consultation. AIPatient Arena provides an EHR-grounded framework for workflow-oriented pre-deployment evaluation of medical LLMs.

Prompt injection defenses evaluated on synthetic benchmarks do not generalize to real enterprise documents, which are longer, denser, and interleave legitimate authority language with factual content. We demonstrate this gap with a real-document benchmark of 122 tasks across five professional domains (financial, legal, medical, scientific, DevOps) using actual SEC filings, Federal Register rules, PubMed abstracts, arXiv papers, and GitHub postmortems. Paraphrasing, the strongest defense on synthetic benchmarks, shows no statistically significant attack success rate reduction on real documents (p=0.500) while degrading utility from 91.8% to 82.8%. We introduce PARSE (Provenance-Aware Retrieval Sanitization), a domain-aware, fact-preserving sanitization pipeline that classifies each sentence by injection likelihood, extracts structured facts before rewriting, and verifies fact preservation via a consistency-checking loop. A directiveness gate routes 59% of real enterprise documents to a lightweight path, concentrating computational cost on high-risk documents. PARSE achieves 15.6% attack success rate -- a 38% reduction versus the 25.4% baseline -- at 86.9% utility, the only condition that is both statistically significant (p=0.014, adequately powered) and maintains near-baseline utility. Practitioners should evaluate defenses on domain-matched real documents, not synthetic proxies.

Membership inference attacks (MIAs) are a canonical way to assess a machine learning model's privacy properties. Although several attempts have been made to evaluate MIAs on language models, the extant literature has suffered numerous difficulties in constructing clean evaluations to test new techniques. In particular, subtle distribution shifts between member and non-member sets can undermine the statistical validity of MIAs; recent work has underscored this by showing that "blind" methods with no access to the underlying model can perform far better than published methods on the same benchmarks. This paper constructs a benchmark for principled evaluation of MIAs against LLMs, by leveraging the insight that training data before and after a fixed point during training are drawn from the same distribution. Therefore, all open-source models with intermediate checkpoints and public training data can be converted into MIA testbeds. We apply our framework to a half-dozen published attacks on the Pythia and OLMo family of models, from 70M to 7B parameters. To facilitate further privacy research, we open-source a modular library for designing and implementing attacks in this setting: https://github.com/safr-ai-lab/pandora_llm.

Vision-Language-Action (VLA) policies have achieved remarkable single-step manipulation, yet they remain brittle precisely where each stage depends on what was just completed. The core issue is structural: short-window VLAs lack an explicit channel for rouxting information across sub-task boundaries, and existing memory-augmented variants either write at every frame, retrieve from demonstration-time stages, or fire at sub-goal events without performing an explicit sub-task-to-sub-task hand-off into the action expert. We identify the sub-goal completion event as the natural temporal unit for cross-subtask memory hand-off, and present WeaveLA (Weave Latent memory for Vision-Language-Action policies), a cross-subtask memory interface that, on top of a frozen VLA backbone, compresses each completed segment into latent tokens via query-driven attention pooling and routes them directly into the action-generation path of the next sub-task. This event-triggered, action-side design preserves the base policy's short-window interface while adding a lightweight cross-subtask channel. Through stratified evaluation on RoboMME with a $π_{0.5}$ backbone, WeaveLA's gains land exactly where the channel is needed: on the hardest repetition slice (SwingXtimes, $N{=}3$), success rises from $0\%$ to $47.8\%$, while single-execution episodes remain unchanged. Per-episode paired analysis confirms the gains are confined to tasks whose causal structure requires cross-subtask information.

Fine-grain clock gating (FGCG) is among the most effective techniques for reducing dynamic power, yet current FGCG optimization flows remain largely manual. Recent LLM-based RTL optimization approaches remain limited by two key drawbacks: (1) the inability to process long waveform traces spanning millions of cycles, and (2) the difficulty of scaling optimization to large hierarchical codebases while preserving correctness. In this work, we present AUTOGATE, the first agentic framework for industry-grade RTL power optimization, enabling workload-aware clock-gating optimization across large hierarchical codebases. AUTOGATE introduces a Machine Learning (ML)-LLM co-design that bridges waveform-level analysis and RTL rewriting. Specifically, we design an ML-based clustering algorithm that distills raw toggling traces into compact, structured representations that guide LLM-based RTL rewriting. This enables accurate identification and application of clock-gating opportunities without requiring LLMs to directly process raw waveform data. To enhance scalability, AUTOGATE employs a hierarchical multi-agent architecture that decomposes large designs into independently optimizable modules, enabling coordinated optimization across deep design hierarchies. We evaluate AUTOGATE on a diverse set of designs ranging from small RTL designs to large industrial-grade codebases. Experimental results show that AUTOGATE consistently reduces dynamic power relative to baselines. Across the small-design suite, AUTOGATE reduces dynamic power by 49.31% on average. On industry-scale designs, it achieves 19.34% and 7.96% dynamic power reductions on NVDLA and BlackParrot, respectively, and up to 6.86% on highly optimized proprietary production designs.

Neural operators are widely used as surrogate solution maps for partial differential equations (PDEs), but full-size models can be costly to store, deploy, and evaluate in many-query scientific workflows. This work introduces Operator Boosting, a stagewise residual-learning framework for constructing compact neural-operator surrogates directly, rather than training a large model and compressing it afterward. Starting from the empirical mean predictor in normalized output coordinates, the method trains a sequence of tiny same-family neural operators on residual fields and incorporates each correction through validation-selected shrinkage. We instantiate the framework with Fourier neural operators (FNOs), DeepONets, and convolutional neural operators (CNOs), and compare boosted tiny stacks against full-size monolithic baselines across one-, two-, and three-dimensional PDE benchmarks from PDEBench, APEBench, and The Well. Across 30 dataset-architecture pairs, 21 show positive mean accuracy gains and 17 have positive confidence intervals, while all boosted stacks reduce trainable parameter count by approximately 72-95%. Best-model comparisons show empirical Pareto improvements on 7 of 10 completed PDE benchmarks, including two-dimensional Navier-Stokes, shallow-water dynamics, Darcy flow, one-dimensional transport and reaction systems, and three-dimensional compressible Navier-Stokes. These results show that Operator Boosting often improves the empirical accuracy-parameter Pareto frontier of neural PDE surrogates, while also exposing PDE- and architecture-dependent regimes where residual boosting fails to offset compression.

Rolling over is one of the earliest milestones in infant motor development, reflecting the emergence of coordinated, whole-body sensorimotor control. Here, we conduct a computational study of infant rolling using MIMo, a virtual infant embodiment equipped with proprioception and vestibular sensation. MIMo learns supine-to-prone rolls with reinforcement learning. Interestingly, the learned behaviors capture developmental trends and coordination patterns consistent with those reported in real infants, including improved performance and faster execution with age. Our results explain how infant capabilities and constraints can give rise to realistic behaviors in artificial agents, with a particular emphasis on how motor development is shaped by the changing body morphology. This work highlights the role of embodied computational models as a powerful tool for studying sensorimotor development.

AI agent performance is not just a modeling problem, it is fundamentally a systems problem. The advanced capabilities of models are realized through agent harnesses. Therefore, a gap between model assumptions and harness behavior can easily prevent the model's full capabilities from translating into agent performance. We formalize this as the `intent-execution' gap: the mismatch between what the model intends and what the harness executes, and vice versa. We argue that minimizing this intent-execution gap is as important as other aspects of harness design such as tools and execution loops. To illustrate the impact of this harness-model alignment, we develop a simple and customizable harness called `Simple Strands Agent' (SSA). SSA aims to find the bulk of common patterns which generalize across different model families (such as Claude, Gemini, GPT, Grok, Qwen), as well as a small number of model-specific preferences. We make two contributions: (i) we $\textbf{reproduce or improve on the pass@1}$ performance reported by diverse model-provider families on popular agentic benchmarks (SWE-Pro, SWE-Verified and Terminal-Bench-2), and (ii) building on an $\textbf{analysis of 138k trajectories generated by SSA}$, we look beyond the $\texttt{pass@1}$ numbers which tend to be relatively even across frontier models. By representing agent trajectories in code state-spaces, we observe model-level differences in problem-solving behavior. Finer-grained metrics such as edit frequency, testing activity, and phase-transitions reveal how individual models allocate effort across different stages of autonomous problem solving.

While Multimodal Retrieval-Augmented Generation (M-RAG) enhances Large Vision-Language Models, it remains highly susceptible to cross-modal hallucinations, causal fabrications, and sycophancy. Furthermore, existing mitigation pipelines often face an intervention paradox: static rules tend to unnecessarily disrupt accurate generations, whereas leaving the multi-modal reasoning completely unguided allows existing mismatches to cascade into severe logical fabrications. To quantify and mitigate these hallucinations, we propose a Multi-Agent system, MODE-RAG, driven by Variational Free Energy (VFE) and internal attention states to dynamically gate interventions. High-risk queries are routed to five stage-specific agents, integrating Monte Carlo Tree Search (MCTS) for rigorous causal derivation and logit perturbations to penalize sycophancy. Dedicated Correction and Overseer agents ensure formatting stability and perform post-hoc factual verification. To objectively evaluate our approach, we introduce ModeVent, a challenging subset derived from the MultiVent dataset. Extensive experiments indicate that our system effectively reduces hallucination rates and logical fabrication, significantly improving the robustness of M-RAG systems.

Simulation enables scalable robot data collection, but raw 3D assets provide only geometry, lacking the semantic, interactive, and physical knowledge needed to specify where and how robots should act. In this work, we present AnnotateAnything, a general automatic annotation framework that converts passive 3D assets into manipulation-ready assets with structured, diverse, and executable manipulation labels. AnnotateAnything is built around two complementary pipelines. First, a unified visual-language annotation pipeline using vision-language reasoning to infer object semantics, interaction constraints, and 3D-grounded cues, providing human-prior guidance for identifying meaningful interaction regions. Second, a fully automatic and massively parallel physics annotation pipeline grounds these priors in each asset's geometry and physical constraints through candidate generation, geometry optimization and trajectory generation. This pipeline produces diverse and executable action annotations, including grasp poses, dexterous contacts, articulation waypoints, insertion directions, hanging affordances, and navigation targets. Using the generated annotations, we further build an asynchronous parallel simulation data-collection system across diverse objects, tasks, and robot embodiments. Experiments demonstrate that AnnotateAnything achieves superior annotation efficiency, data-collection efficiency, and task success rates over existing annotation and data-generation pipelines, while also supporting downstream tasks such as affordance detection, robotic VQA, and visual instruction finetuning. We provide project materials on the project page and plan to release the full code, annotations, and benchmark to facilitate future research. Videos, code, demo assets, and annotations are provided in supplementary materials Project page: https://tourmaline-caramel-169490.netlify.app.

Large language models (LLMs) are becoming a major way for consumers to find products, but we do not yet understand how brands compete in this new channel. We study brand dynamics in LLM recommendations using skincare products -- a category where consumers cannot easily judge quality before buying and must rely on brand reputation -- across three commercial LLMs (GPT-4o-mini, Claude Sonnet, Gemini 3 Flash), with a robustness check on search goods. In three experiments, we find: (1) a Conditional Monopoly where well-known brands get recommended 100% of the time (IAI = 10.0) when all products have the same specifications, but this dominance disappears with less than a +0.1-star rating advantage for a competitor; (2) authority-style marketing language, including fabricated clinical-evidence claims, breaks this monopoly at a Bias Surplus Value equal to +0.17 rating points, with each model responding differently; and (3) a social dilemma in multi-brand GEO competition: when all brands adopt the same optimization strategy, individual payoff falls from +0.802 to +0.007 in our payoff proxy, and non-participating brands receive zero recommendations in our tests. Our results suggest that generative engine optimization (GEO) should be studied not only as a security risk, but also as an emerging marketing practice that shapes market competition.

This paper presents a contact-based 3-D surface measurement method based on a Digital Fringe Projection (DFP) system, belonging to the vision-based tactile sensing family pioneered by the commercially successful GelSight sensor. Such sensors have proven effective for robotic fingertip manipulation and contact sensing. However, because GelSight employs photometric stereo with RGB LEDs, it does not measure absolute depth directly but instead infers it by integrating estimated surface gradients, which can accumulate reconstruction errors; in addition, it becomes increasingly difficult to calibrate as the sensing area grows, and its depth accuracy is challenged on highly reflective or transparent objects. To overcome these drawbacks, we propose a fringe-projection-based contact measurement technique that performs triangulation-based 3-D reconstruction on a coated silicone contact surface, providing dense per-pixel surface geometry and full-field 3-D shape measurement over the contact region. By integrating high-accuracy digital fringe projection into the sensor, our approach simplifies calibration over larger areas and enhances depth precision for complex surfaces. Experimental results, including a direct comparison with a GelSight Mini sensor, a sphere-fitting accuracy evaluation, and an uncertainty analysis, confirm that the proposed method significantly improves the accuracy and stability of structured-light-based 3-D measurements, allowing reliable reconstruction of objects with diverse optical properties.

Automated classification of standard echocardiographic views is crucial for efficient clinical workflow but faces three main challenges. First, publicly available datasets are scarce and limited in scale and view coverage. Second, the performance of some modern video-level architectures for echocardiographic view classification remains underexplored. Third, some view categories exhibit highly similar spatial appearances, making single-frame features insufficient for discrimination, while heterogeneous frame quality complicates robust temporal information fusion. To address these challenges, we release the Echocardiographic Videos of Nine Views (EV9V) dataset, comprising 5,138 videos, 910,579 frames, and 9 standard views, which is, to the best of our knowledge, the largest publicly available echocardiography video dataset. Using EV9V, we systematically benchmark representative video classification architectures, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Transformers. Furthermore, we propose a Spatio-Temporal Fusion Model (STFM), an efficient dual-stream CNN-LSTM (Long Short-Term Memory) framework that jointly captures spatial anatomical structures and temporal cardiac dynamics. The proposed framework leverages uncertainty-aware learning to preferentially sample representative video segments during training and evidence-based fusion during inference, improving robustness to variations in frame quality across echocardiographic videos. Extensive experiments demonstrate that our method achieves competitive performance across diverse video classification models, validating the effectiveness of uncertainty-aware spatio-temporal learning for echocardiographic view classification. The code is available at https://github.com/bgx666/stfm.

Large Vision Language Models (VLMs) excel at visual question answering and semantic grounding, but their capacity for autonomous logical reasoning remains underexplored. Existing anomaly benchmarks emphasize visual errors or direct prompting rather than the physical and social common sense needed for open-world deployment. To address this, we introduce LAD-bench, a benchmark of more than 1,000 curated synthetic images with logical anomalies across four domains: Residential, Urban, Collaborative, and Nature. We further propose a Tiered Prompting Protocol based on progressive disclosure, which measures how much explicit assistance a model needs to localize and reason about a logical fault. Evaluating leading foundation models reveals substantial weaknesses: even the best achieves only 70.11% overall accuracy, showing that implicit logical fault detection remains unsolved. Crucially, models often fail to identify anomalies even after receiving explicit hints in deeper tiers. By surfacing these limitations in sequential multimodal reasoning, LAD-Bench offers a rigorous framework for advancing the safety, reliability, and cognitive alignment of autonomous visual systems. Dataset and Code: https://huggingface.co/datasets/SahasraK/LADBench

Generative AI has advanced the ability to render photorealistic or artistic images, yet it remains limited in a key aspect of human creativity: interpreting ambiguous shapes. This phenomenon, rooted in pareidolia, allows humans to perceive meaningful forms in random patterns such as clouds, stones, or leaves. To computationally replicate this imaginative process, we introduce Visual Retrieval-Augmented Generation (Visual-RAG), a framework that generates animal art directly from natural silhouettes. Our method retrieves structurally similar animal shapes from a curated corpus of 28,586 high-quality silhouettes and uses them as reference exemplars to guide diffusion-based generation with ControlNet and IP-Adapter. Ablation studies confirm that shape Context with RANSAC provides the most accurate alignment, while removing shape standardization reduces the inlier ratio to just 13.4\%, underscoring the importance of structural fidelity in Visual-RAG. A user study with 12 participants evaluated the outputs in terms of aesthetics, silhouette fidelity, and overall impression. Results reveal that while Visual-RAG provides plausible interpretations, challenges remain in achieving high perceptual impact. This work lays the foundation for computational pareidolia, showing how machines can contribute to the early stages of imaginative discovery.

Event-enriched image captioning describes not only visible content but also the broader context of events, including timing, location, and participants, capabilities missing in most pixel-bound models. We propose the Contextual Image-Article Narrator (CIAN), a multi-stage framework that enriches captions with external narratives. CIAN retrieves relevant articles using SigLIP, summarizes them to guide a Narrative Generation stage with a LoRA-fine-tuned Qwen model, and applies N-Gram-based Refinement for fluency and coherence. On the OpenEvents-V1 benchmark, CIAN achieves high retrieval performance (mAP 0.979) and improves caption quality, increasing CIDEr from 0.030 to 0.094. These results highlight the effectiveness of retrieval-augmented reasoning combined with linguistic refinement for generating context-aware, human-like captions.

Dexterous robotic hands face a longstanding trade-off among dexterity, compactness, and affordability. Particularly, high-degree-of-freedom designs typically demand complex actuation and transmission, hindering integration into human-scale forms. To address these challenges, this work presents a compact, low-cost linkage-driven anthropomorphic hand that achieves high dexterity, structural integration, and human-hand-like functionality. The hand integrates 20 joints driven by 16 independent actuators, with all actuation, sensing, and transmission components compactly embedded within a human-hand-sized structure. The resulting prototype weighs only 320g at a total cost below USD 400. To meet these objectives, a hybrid mechanical architecture combining planar and spatial linkage mechanisms is proposed, enabling decoupled multidirectional motion, biomimetic joint synergies, and high passive load-bearing capability. The thumb further incorporates biomimetic features supporting human-like reconfiguration and opposition movements. Through the coordinated integration of these mechanisms and structural layout, the prototype achieves a highly integrated design with anthropomorphic dexterity. Experimental evaluations demonstrate that the hand achieves the maximum Kapandji score, reproduces all 33 Feix grasp types, and performs stable grasping and dexterous manipulation across a wide variety of daily objects and tools. These results validate the proposed hand as an affordable, compact, and mechanically efficient platform for dexterous manipulation, teleoperation, and robot learning in human-centered environments.

Pathological images are inherently multi-scale, requiring pathologists to integrate evidence from global tissue architecture at low magnification to cellular morphology at higher magnification for accurate diagnosis. While existing pathological datasets for vision-language model (VLM) include various scales, they often lack an explicit cross-scale reasoning objective. This limitation prevents VLMs from capturing essential cross-scale representations and learning evidence-based reasoning. To bridge this gap, we introduce the first cross-scale training and evaluation paradigm that formulates pathology interpretation as multi-magnification reasoning. However, creating such a task reveals a critical challenge: multi-image visual question answering (VQA) is prone to text-only shortcuts, which allow models to guess answers using magnification-dependent artifacts rather than visual evidence. To address this, we propose a leakage-aware curation pipeline that combines adversarial text-only screening with constraint-guided question design. Using this pipeline, we construct Scale-VQA, a high-quality benchmark with 4,685 multiple-choice questions grounded in 2,537 pathology images across multiple magnification levels. Finally, we present ScaleReasoner-R1, a model trained via reinforcement learning to optimize performance on the cross-scale VQA task. ScaleReasoner-R1 achieves state-of-the-art performance on our cross-scale reasoning benchmark and generalizes to SOTA performance on established single-scale benchmarks. Findings suggest that even the limited cross-scale supervision can significantly improve pathological understanding. The code and demos will be open-sourced.

Generative robot policies typically begin action generation from an observation-independent standard Gaussian distribution, leaving the choice of source distribution underexplored. This work asks a simple question: where should action generation begin? We propose LeaP, a Learnable source Prior that replaces the standard Gaussian with a proprioception-conditioned diagonal Gaussian over action chunks. Parameterized by a lightweight MLP, LeaP jointly predicts the mean and state-adaptive variance of the source distribution, while keeping the downstream generator architecture and inference solver unchanged. This design provides an observation-informed yet stochastic initialization, allowing the generator to focus on precise action refinement rather than transporting samples from an uninformed noise source. On 15 RoboTwin manipulation tasks, LeaP achieves an average success rate of 81.6%, outperforming four representative baselines -- including deterministic-source methods, a no-prior counterpart, and a diffusion-bridge policy -- by 6.5 to 25.5 percentage points. The same prior consistently improves both flow-matching and diffusion-bridge generators, while using fewer parameters and converging faster. The advantage carries over to real-world deployment, where LeaP attains the best performance. These results suggest that the source distribution is an independent and reusable design axis for generative robot policies, complementary to the choice of generative dynamics.

Long-form serialized audio drama, with arcs that run for 200 to 800 episodes, is a major creative medium and a setting where frontier large language models (LLMs) fail. We benchmark 21 models, spanning classical, fine-tuned, open-frontier, closed-frontier, and reasoning tiers, on a uniform set of structural narrative metrics. All closed-frontier systems saturate at a plot-beat F1 in the band [0.78, 0.81] and collapse by about -0.20 F1 at horizon h=200. We introduce NarrativeWorldBench, an open benchmark of nine narrative-structure metrics evaluated across horizons h in {10, 20, 50, 100, 200}, with cross-lingual evaluation across four Indic languages (Hindi, Tamil, Telugu, Marathi). We introduce N-VSSM, a Narrative Variational State-Space Model that maintains a structured 256-dimensional latent world state over more than 200 episodes via a Mamba-2 backbone with an event-conditioned posterior and an 8B decoder. N-VSSM holds plot-beat F1 >= 0.84 across all horizons at 4x lower compute than the closed-frontier band. A learned Cultural Transfer Function lifts cross-language fidelity by +0.20 to +0.23 Likert points. In a within-subjects writer study (n = 12 professional authors, 240 trials), N-VSSM is preferred over Claude Opus 4.5 on long-arc consistency 71% of the time and rated +1.3 Likert points higher on controllability.

In this paper, we generalize the concept of Voronoi regions, define agent utility as the integral of a utility density over the corresponding Voronoi region, derive gradients of the utility, and illustrate the approach in a two-team example from soccer. The generalization of Voronoi regions is in the form of so-called Cost-Induced Voronoi (CIV) regions, where the agent state space may differ from the space being partitioned. One example of such regions is when the cost is given by the optimal solution of an LQR control problem. Then the agent states include position as well as velocity, while the partitioned space only includes positions. The agent utility is defined by integrating some utility density over the CIV region of the agent. This utility density might be the probability density of some beneficial event, such as receiving a pass in soccer. The utility is then the overall probability of receiving a pass and the gradient represents a way to improve that probability. We show how this utility gradient can be computed using the Reynolds Transport Theorem from fluid mechanics, and that this approach achieves similar accuracy while reducing computation time by about an order of magnitude compared to a baseline finite-difference approximation.

End-to-end autonomous driving has achieved state-of-the-art performance on benchmarks and real-world deployments. Its standard training recipe, however, is expensive across all stages: collecting and labeling millions of driving frames is costly, and closed-loop RL on images is bottlenecked by the per-step cost of photorealistic rendering plus a forward pass through a large vision backbone. Self-play in vectorized simulators changes the economics: millions of rollout steps per second, and a state distribution naturally rich in collisions, near-misses, and recoveries that no driving log contains. Our approach exploits this asymmetry by decoupling learning to drive from learning to see. We pretrain a single policy by self-play, then align its latent space with a pretrained vision backbone, through the action KL divergence and a batch-relational low-rank structural loss. The action target comes from the self-play policy, so alignment never supervises against a logged trajectory: a paired dataset of (image, scene-state) frames suffices, with no need for the curated expert demonstrations that imitation pretraining is built on. On photorealistic 3D Gaussian splatting closed-loop scenarios, the resulting end-to-end policy matches or exceeds prior end-to-end methods.

Internet videos constitute the largest reservoir of embodied human manipulation knowledge, yet converting arbitrary RGB footage into actionable robot training data remains a major bottleneck. Existing lab- or factory-collected datasets are narrow in scale and diversity, limiting open-world robot learning. Instead of proposing a static dataset, we introduce EgoInfinity, a universal 4D hand-object interaction data engine that enables web-scale data generation for robot retargeting and learning. EgoInfinity is a modular engine integrating perception, segmentation, reconstruction, interaction-aware refinement, and retargeting to automate this traditionally unscalable video-to-action problem without human-in-the-loop annotation. Its modular design lets the engine continuously benefit from advances in any incorporated component. With EgoInfinity, in-the-wild human manipulation videos are lifted into agent-agnostic, metric 4D hand-object representations, including hand trajectories, 6-DoF object poses, and contact-relevant states. Rather than naively connecting standalone components, EgoInfinity combines cross-module metric calibration with interaction-aware refinement to improve physical reliability, reducing drift and contact inconsistencies common in pure visual reconstruction. We further propose a novel motion retargeter that compiles the recovered 3D hand motions into executable joint trajectories for diverse robot morphologies, enabling video-to-action retargeting on any robot from arbitrary viewpoints and shot sizes (e.g., the human body is only partially visible). We validate EgoInfinity across perception fidelity, kinematic feasibility, contact consistency, cross-embodiment generalization, and real-robot skill acquisition (e.g., grasping, cutting, wiping, and pouring), demonstrating a scalable bridge from internet videos to executable robot behavior for open-world robot learning.

Understanding hands and the objects they interact with, both directly and through tools, is a key step for tasks ranging from action perception to 3D reconstruction and robotics. Our paper provides several contributions to the Hand-Object Interaction (HOI) understanding literature: (1) HOI-DETR, a new framework that introduces hand-object and object-object interactions to the Co-DETR architecture to produce a state-of-the-art method; (2) a comprehensive HOI evaluation suite of 4 diverse datasets, including a video benchmark derived from the HD-EPIC dataset and fresh annotations that improve the Hands23 benchmark and (3) a trained checkpoint that significantly improves the state of the art across Hands23, HOIST, FineBio, and HD-EPIC, including mAP gains of over 20 percentage points on Hands23 and FineBio. Our ablations confirm the contributions of each model component.

Respiratory-rate (RR) monitoring is a critical component of remote triage and victim assessment in emergency response, disaster recovery, and infectious-disease scenarios, where minimizing physical contact can reduce responder risk and improve operational safety. However, field deployment of contactless RR monitoring remains challenging due to variable illumination, posture changes, platform heterogeneity, and the impracticality of wearable sensors in hazardous environments. In this paper, we present a modality-adaptive contactless RR monitoring framework for heterogeneous mobile robots with onboard edge computing. The proposed system combines brightness-adaptive sensor selection across RGB, thermal, near-infrared (NIR), and low-light cameras, keypoint-guided chest ROI extraction for posture-robust monitoring, and a signal-quality-index (SQI)-based filtering mechanism for reliable respiratory estimation. We implement and evaluate the framework on three robotic platforms spanning quadruped and wheeled locomotion and multiple edge-computing architectures. Experiments conducted across diverse lighting conditions, subject poses, and robot-to-subject distances demonstrate that the framework generalizes across platforms without per-platform algorithmic retuning, while revealing modality-specific operational boundaries. RGB provides the broadest coverage up to 8m, NIR remains effective up to 6m, thermal is reliable only at short range, and low-light sensing supports monitoring in complete darkness up to 8m. Overall, the results demonstrate the feasibility of multimodal contactless RR monitoring on mobile robots and support its use as a foundation for autonomous triage and victim assessment in hazardous search-and-rescue settings.

Autonomous driving has shifted towards end-to-end policy learning, where reliable, interpretable policy evaluation is a fundamental challenge as driving quality is highly context-dependent. Commonly used rule-based driving metrics like EPDMS are interpretable but lack context-awareness, while recent VLMbased evaluations are context-aware but limited by ambiguous VLM outputs and weak physical grounding. To evaluate driving in a manner that is both interpretable and context-aware, we introduce DriveJudge. DriveJudge is a driving evaluation agent that combines rule-grounded evaluation with Vision-Language Model (VLM) reasoning and selectively invokes physically-grounded deterministic rule functions after interpreting the environmental context. To train and evaluate DriveJudge, we curate a large-scale dataset of 33,577 challenging driving samples with human annotations on whether the driving behavior is reasonable in the given scenario. With this dataset, we address the underexplored problem of driving metric evaluation, and introduce two human-aligned benchmark tasks: Driving Quality Classification and Trajectory Preference Selection. DriveJudge outperforms EPDMS for driving quality classification by 21.23 AUC, and the recent VLM-based DriveCritic for trajectory preference selection by 6.5%, setting a new standard for interpretable and precise driving evaluation.

Untranslatability, cases where meaning cannot be directly preserved across languages, is well-studied in linguistics but underexplored in NLP. As machine translation (MT) systems improve on standard benchmarks, their limitations increasingly concentrate in such cases, where translation cannot be reduced to one-to-one equivalence. We introduce a structured ontology of untranslatability along with a taxonomy of compensation strategies, which are specific techniques to convey meaning under these untranslatable circumstances. We operationalize this framework into a multilingual dataset of untranslatable sentences paired with strategy-based translations, enabling controlled analysis of translation behavior. Initial human preference studies suggest that translation quality depends on the strategy used, with consistent preferences for outputs that include explanatory context, known as the Annotation compensation strategy. Our framework and dataset provide a foundation for studying and modeling strategy-informed machine translation.

Speech offers a uniquely informative window into health by simultaneously engaging neurological, motor, respiratory, and vocal systems. Current clinical speech AI methods have largely progressed through isolated condition-specific studies, making results difficult to compare and generalization difficult to assess. We introduce SpeechDx, a large-scale benchmark for clinical speech AI spanning 12 datasets and 27 tasks across diverse health conditions. To enable evaluation across shared clinical mechanisms, SpeechDx structures tasks by the stage of speech production they disrupt: conceptualization, formulation, and articulation. The benchmark tests generalization by including tasks with limited labeled data and evaluating the same health condition across multiple datasets, distinguishing clinically meaningful patterns from dataset artefacts. We systematically evaluate 12 state-of-the-art audio encoders across all tasks and under zero-shot cross-condition transfer. Results show that large-scale speech models represent the strongest overall baselines, domain-specific models improve performance only on closely matched tasks, and no current representation generalizes reliably across the clinical speech landscape. SpeechDx establishes a shared evaluation framework for tracking progress toward general-purpose clinical speech representations

Event cameras are bio-inspired sensors that asynchronously capture logarithmic intensity changes, offering inherent advantages in high-speed and high-dynamic-range scenarios. However, the sparse and asynchronous nature of event streams poses a fundamental challenge for modern deep learning architectures. To enable compatibility with standard models, most existing approaches partition the accumulation window into fixed temporal sub-bins. While effective for spatial processing, this internal discretization discards fine-grained temporal structure and constrains inference to the low temporal frequencies imposed by training supervision. To address this limitation, we propose FATE, a unified framework built upon a novel Pillar Encoding (PE). While operating over discrete macro-accumulation windows dictated by the target frequency, PE avoids internal temporal sub-binning. It organizes events into spatial pillars and approximates their intra-window evolution via projection onto a continuous-time orthogonal polynomial basis. This formulation yields an L2-optimal representation that retains rich temporal dynamics in a dense pseudo-image, mitigating information loss under sparse event conditions. To fully leverage this representation, we introduce Frequency-Aware Training (FAT), a soft mean-teacher curriculum that generates temporally dense pseudo-labels, effectively bridging the mismatch between low-frequency supervision and high-frequency inference. Extensive experiments demonstrate that FATE generalizes across architectural paradigms and consistently outperforms strong baselines. It enables robust object detection at high temporal resolutions up to 200 Hz, while incurring minimal overhead in parameter count and inference latency

Real-time trajectory generation for on-orbit robotic servicing is challenging due to the nonlinear coupling between spacecraft bus motion, manipulator dynamics, visibility cone, and trajectory-level safety constraints. This paper studies learning-based warm-starting for sequential convex programming (SCP) in the terminal approach of a space manipulator toward a tumbling target. The proposed framework decomposes the problem into a system center-of-mass translational planning stage and a coupled attitude--manipulator torque-allocation stage, and applies a causal transformer warm-start to the latter, which constitutes the dominant computational bottleneck. Linear and flow matching action decoders are compared under different action-chunking and training dataset sizes, and the resulting warm-starts are evaluated under both cost-optimal and feasibility projection using SCP. Across 300 held-out scenarios, the learned warm-start reduces the second-stage SCP iteration count by up to 28% and the runtime by 23% while preserving the final control-cost distribution. When the learned warm-starts are used for nonconvex feasibility projection, they nearly halve the runtime relative to cost-optimal SCP, while avoiding the catastrophic high-cost tail behavior observed when initialized heuristically. These results indicate that sequence-model warm-starts can improve both the computational efficiency and trajectory robustness of optimization-based terminal guidance for space manipulation.

Robotic assistance in household environments requires not only predicting where objects should be placed, but also reasoning about when objects should not be placed at all. Existing approaches to personalized object rearrangement primarily focus on placement decisions under the assumption of clean observations and complete actionability, limiting their applicability in realistic, cluttered, and partially erroneous settings. In this paper, we introduce APOLLO, a hybrid framework for abstention-aware personalized object rearrangement that combines a lightweight, personalized embedding model (PEM) with selective large language model (LLM) assistance. PEM is trained for each user-environment pair using a small number of demonstrations, operates entirely on CPU, and produces uncertainty estimates, which are used to selectively invoke LLM-based reasoning only for ambiguous decisions, balancing efficiency, privacy, and reasoning capability. To evaluate this formulation beyond existing benchmarks, we introduce APOR, a synthetic, LLM-generated dataset that captures room-level, multi-furniture environments, diverse organizational profiles, explicit abstention behavior, and noisy partial scene context. Extensive experiments on both PARSEC and APOR provide initial evidence that APOLLO improves over prior LLM-based baselines in controlled benchmark settings while substantially reducing LLM usage. Code is available at https://github.com/PaInt-Lab/APOLLO.

Text-to-video retrieval in operating rooms (OR) is an enabling technology for OR safety, as it allows stakeholders to retrieve and inspect recordings of specific events. However, because the most safety-critical events may not follow the common structure, to unlock its full potential text-to-video retrieval must be able to handle implicit queries that require reasoning to identify the right video (e.g., the step right before clipping). However, existing methods rely on global embeddings that cannot reason over such queries. We propose OR3, a text-to-video retrieval method that converts clips into action-driven digital twins (ActDTs), grouping concurrent subject-action-object triplets under non-overlapping temporal intervals. Moreover, rather than cross-modal matching through paired encoders, OR3 performs imagination-based retrieval where an LLM generates hypothetical ActDTs from queries. This enables intra-modal matching via a single encoder trained with ActDT-tailored hard negatives. Finally, evidence-grounded refinement revises imagined ActDTs based on discrepancies with top candidates to capture procedure-specific patterns. We construct a benchmark from MM-OR with 276 implicit queries across four reasoning categories over 386 clips from robotic knee procedures. OR3 achieves 57.6 R@1 and 77.3 R@5, outperforming the strongest baseline. These results demonstrate that OR3 enables fine-grained discrimination between visually similar OR video clips through temporal action reasoning.

Unified multimodal models (UMMs) have recently emerged as a promising paradigm for integrating multimodal understanding and generation within a single autoregressive transformer. However, during multimodal instruction tuning, these models often exhibit pronounced modality imbalance: language gradients dominate optimization, thus leading to lower image generation quality, especially under parameter-efficient fine-tuning such as LoRA. In this work, we systematically analyze modality imbalance in LoRA-based fine-tuning of UMMs for interleaved text-image generation. We show that vision modality performance degrades substantially more than text modality performance when compared to unimodal counterparts, and that modality-specific gradients can differ by orders of magnitude across various tasks and layers. Motivated by this observation, we reformulate the multimodal instruction tuning as a bi-objective optimization problem and propose Pareto LoRA, a Pareto-optimal gradient integration strategy that balances the text and image objectives by modulating the gradient direction and strength. Experiments on the CoMM benchmark with Emu2 demonstrate that Pareto LoRA consistently improves multimodal generation balance, achieving up to 44.9% gains in perceptual image quality over vanilla LoRA while maintaining comparable text performance.

Production planning increasingly has to treat workforce capability as a decision variable: certifications lapse when skills are not maintained, new products require skills the current workforce does not hold, and reskilling competes for the same worker hours needed for production. Existing operations benchmarks usually treat labor as exogenous, while workforce-planning models with skills and learning are rarely released as reusable testbeds. We introduce SkillChain-Gym, a benchmark specification for reskilling-aware production-inventory control: a single-site environment with stylized worker skill-state dynamics, hard threshold certification, forgetting, and capacity-consuming training actions constrained by the same per-worker time budget as production. The benchmark includes seed-controlled disruption scenarios, three feasibility modes with projection diagnostics, deterministic replay, and metrics covering operations, resilience, capability growth, and training-access distribution. We evaluate production-only, reactive adaptive, water-filling adaptive, and static-insurance policies with budget variants over 60-shift horizons with paired statistical tests. The results are regime-dependent rather than a ranking. Training-capable policies dominate the production-only baseline, and maintenance training is necessary under forgetting even without disruptions. Among training-capable classes, adaptive training helps when bottlenecks are visible in the forecast, while a lean static cross-training plan, a deliberately favorable comparator whose structure encodes relevant skill contingencies, acts as strong insurance under surprise shocks and absenteeism. Capacity slack and the forgetting rate govern the boundary between these regimes. No policy class dominates across regimes, motivating forecast-driven controllers that decide when to buy skill insurance and when to react.

Existing vision encoders for robotics face a fundamental bottleneck: robotic datasets lack the scale necessary for large-scale pre-training. Prior work circumvents this data scarcity by turning to internet-scale image and language data or egocentric human video. While these models show promise, neither paradigm learns from paired vision and action data, which downstream visuomotor control policies require. However, robot trajectories, the most direct source of this paired signal, are not available at pre-training scale, motivating us to extract action signals from abundant human video instead. To this end, we introduce CAIP (Contrastive Action-Image Pre-training), a vision encoder that treats human hand poses from large-scale egocentric video as a proxy for end-effector actions. By extracting 3D hand keypoints, a representation that aligns naturally with downstream robot action spaces, CAIP learns a unified action-image representation through a contrastive objective. Leveraging 32,041 hours of egocentric human video and only 88 hours of robotic manipulation data, CAIP outperforms state-of-the-art vision encoders including DINOv2, SigLIP, MVP, and R3M. Evaluated on a challenging real-world dexterous manipulation setup using Dexmate Vega and Sharpa Wave hands, CAIP yields performance gains of more than 30% on tasks involving folding, pouring, and fine-grained manipulation. Our results show that our method of contrastive action-centric pre-training yields a scalable path to achieving robust visual representations better suited for physical interaction.

This work describes the participation of the MLLP-VRAIN research group in the shared task of the IWSLT 2026 Simultaneous Speech Translation track. Our submission utilizes the recently released Parakeet and Qwen 3.5 models to create a robust, cascaded solution for long-form SimulST through the use of adaptive "black-box" policies. We explore relaxations of these policies to achieve better quality-latency trade-offs. Compared to last year, we participate on all language directions. In addition to this, for the En$\rightarrow${De, It, Zh} directions we also participate in this year's new context track employing a combination of ASR word-boosting and a RAG mechanism of offline pre-translated exemplars to guide generation and enrich our system with domain-specific context. Finally, we provide a detailed latency analysis of our system. Compared to last year, results on the MCIF En$\rightarrow$De test set shows a substantial quality improvement of +5.82 XCOMET-XL. Our context track processing further improves performance by +1.03.

Reinforcement learning (RL) has become a central paradigm for post-training large language models. Existing critic-free RL methods typically generate a group of rollouts for the same question to estimate value baselines for advantage computation. However, this design suffers from data inefficiency, group synchronization barriers, and inflexibility with structured rollouts. In this work, we revisit the role of the ``group'' and show that its underlying function is not merely to estimate baselines but to prevent false penalties on negative samples. Building on this insight, we propose negative token filtering, a simple and effective strategy that enables stable single-rollout training. We apply it to two batch-level advantage methods, achieving comparable performance on reasoning tasks and stronger performance on agentic tasks relative to group-based RL techniques.

Remote-sensing vision-language models (RS-VLMs) have advanced Earth-observation analysis toward visual interpretation and instruction-following, yet fall short of operational geo-intelligence, which demands tool-grounded spatial reasoning and structured, evidence-backed decisions. We introduce GeoDisaster, an operational geospatial disaster reasoning benchmark with 2,921 verified instances across 43 question types and five task families: deforestation monitoring, multi-hazard analysis, building-damage assessment, flood-safe routing, and Sentinel-1 SAR flood monitoring. Instances integrate heterogeneous EO/GIS evidence-optical and SAR imagery, raster masks, vector geometries, road networks, and exposure layers-spanning hazard detection, damage assessment, exposure estimation, and diagnostic report generation. Ground-truth answers are grounded in executable geospatial workflows and deterministic consistency checks, removing the need for language-model annotation. We further propose an orchestrated multi-agent framework with 18 disaster-oriented tools, where role-specialized agents coordinate through explicit execution contracts, aligned via Role-Contract Expectation Alignment (RCEA): failure-aware supervised fine-tuning combined with contract-grounded reinforcement learning over dense step-level signals. Experiments show that GeoDisaster challenges existing RS-VLMs and agentic systems, while RCEA improves tool use, evidence grounding, state consistency, and decision generation.

Continuous AI inference on resource-constrained edge hardware introduces deployment effects that are largely invisible to conventional benchmark evaluation, including temporal instability in streaming video, thermal throttling under sustained load, and workload-dependent performance variability. We present Edge-TSR, a deployment-oriented continuous edge inference system for sustained roadside perception on the NVIDIA Jetson Orin Nano. Edge-TSR integrates detection, tracking, fine-grained classification, and a lightweight track-aware temporal stabilization mechanism that improves streaming inference consistency with negligible computational overhead. Our central finding is that benchmark-centric evaluation systematically overstates deployed edge inference performance. Across three state-of-the-art baselines, we observe consistent 20-30% relative degradation when transitioning from static-image evaluation to real-world streaming deployment. Edge-TSR addresses this gap through temporal inference stabilization, recovering up to 10.16% classification accuracy over per-frame inference baselines while maintaining sustained real-time performance under continuous operation. We evaluate the complete system under diverse real-world deployment conditions, jointly characterizing inference quality, latency, throughput, and thermal behavior during long-duration operation. A 55-minute vehicular deployment over a 26 km route demonstrates sustained operation at 16.18 FPS within safe thermal limits on a single embedded device without cloud offload. Our findings show that deployment-aware evaluation and temporal inference stabilization are necessary components of continuously operating edge AI systems intended for real-world sensing deployments. We release a sample annotated streaming video evaluation dataset and full system implementation to support reproducible deployment-centric evaluation.

A model that lies while knowing the truth is the central case ELK cannot handle with behavioral evaluation alone. We ask whether such deception leaves an internal signature distinguishing it from honest error. Our key move is a control for wrongness: we contrast a sleeper agent (knows the truth, lies on trigger) against a naive liar (fine-tuned to emit the same wrong answers with no honest training). Both produce identical wrong outputs; any difference is about knowledge conflict, not incorrectness. We find deceptive forward passes carry a conflict signature - 2.1-2.3x higher residual rank than naive-liar passes on the same wrong answer - strong enough to identify which of two responses is the lie with 100% accuracy and no labels, across GPT-2 small/medium (three seeds) and three instruct models. Across Qwen2.5-1.5B/7B and Phi-3-mini, instructed deception raises residual rank on every tested fact (18/18, 40/40, 34/34); on Phi-3, lies separate perfectly from both honest answers and hallucinations (AUC 1.0, Wilcoxon p~6e-11). The signature survives strategic self-constructed deception (model invents its own lie, AUC 1.0), active concealment attempts (AUC 1.0), and length-controlled replication (20/20, AUC 1.0, p~1e-6). Using basis-free relative representations, a probe trained on one model family detects deception in two other families zero-shot (mean AUC 0.933), surviving simultaneous architecture and format change (AUC 0.821), and transfers across five languages (AUC 1.000, length-controlled). The signature is read-only: detectable but not injectable (0/8 both directions). Honest limitations and six negative experiments are documented in full.

This paper develops and analyses the Fair Play Automatic Market Maker (FP-AMM), a programmable electricity allocation mechanism in which scarcity allocation is treated as a controlled, stateful, and auditable cyber-physical process. Existing mechanisms such as locational marginal pricing are memoryless and cannot account for historical service outcomes, preventing guarantees of equitable treatment across market intervals. The FP-AMM employs a two-stage stochastic clearing rule comprising service-priority sampling and inverse-fairness weighting, coupled with a DC-OPF feasibility set and bounded shortage memory updated through a saturated integrator. Four main results are established. First, the shortage-memory state is invariant in $[0,1]^N$ and the update map is a contraction with rate $1-β$. Second, the intra-interval clearing operator converges linearly to a unique fixed point with contraction factor $q\in(0,1)$. Third, under the Fair Play priority rule, the per-node delivery ratio converges almost surely to the contracted target $F^\star$, with a finite-time $O(1/\sqrt{T})$ bound obtained via Lyapunov analysis of the deficit recursion. Fourth, event-triggered execution guarantees practical ultimate boundedness of the allocation tracking error and quantifies the computation-fidelity trade-off. The mechanism is validated on the IEEE 14-, 57-, and 118-bus systems over $T=5000$ market intervals. Fairness convergence to $F^\star$ is achieved on all benchmarks, peak weak-bus fairness error is reduced by 54% on the IEEE-57 network and by up to 55% relative to an equal-weight baseline during scarcity periods, and DC feasibility is maintained throughout.

Vision-Language-Action (VLA) models benefit from large-scale and diverse embodied data, yet scaling robot trajectory collection is costly and labor-intensive. Recent advances show that large-scale egocentric human videos provide complementary real-world supervision in pretraining. However, joint training on human and robot data remains challenging due to divergences in action spaces, embodiment structures, temporal dynamics, and supervision quality. We introduce ACE-EGO-0, a unified VLA pretraining framework jointly leveraging heterogeneous data sources. To extract large-scale pretraining supervision from egocentric human videos, we build a scalable egocentric video-to-action pipeline that converts raw human videos into robot-format pseudo-action trajectories. To make these labels comparable with robot demonstrations, ACE-EGO-0 uses a unified action representation based on camera-space actions, morphology conditioning, and time-aligned action chunking. To robustly leverage noisy pseudo-action supervision from egocentric human videos, we formulate a reliability-aware training objective with a human auxiliary loss that concentrates supervision on reliable signals. We instantiate ACE-EGO-0 on 4.53K hours of robot and simulation data, together with 1.48K hours of pseudo-action-labeled egocentric human data. Experiments show that incorporating large-scale human supervision under reliability-aware weighting consistently improves both unified joint pretraining and supervised fine-tuning. ACE-EGO-0 achieves state-of-the-art performance on RoboCasa GR1 TableTop and RoboTwin 2.0, while demonstrating strong transfer to real-world bimanual manipulation.

Current multilingual evaluations for Vision-Language Models (VLMs) assume a one-to-one mapping between language and orthography, overlooking billions of users of multi-script languages. We introduce PuMVR (Punjabi Multimodal Visual Reasoning), a benchmark of 1,000 strictly parallel image-text instances across Punjabi's three active scripts: Gurmukhi, Shahmukhi, and Roman. Evaluating 10 state-of-the-art VLMs, we expose a substantial and systematic Script Gap. Models frequently solve visual tasks in one script while failing identical tasks in another, with accuracy deltas reaching 16%. Crucially, visual input boosts absolute performance uniformly yet does not close the orthographic gap. Furthermore, cross-script in-context transfer is highly brittle, exposing script-locked knowledge representation. Supported by McNemar tests across all script pairs, our findings demonstrate that current "multilingual" VLMs are not truly multi-script. We propose the Script Consistency Rate (SCR), which falls as low as 24.8% on our benchmark, as a mandatory metric for script-agnostic evaluation to ensure equitable AI access. Data and code are available at: https://github.com/prabhjotschugh/Not-Truly-Multilingual-PuMVR.

Answering navigation-relevant questions over long egocentric videos requires retrieving and organizing evidence distributed across distant temporal moments while maintaining spatial and contextual consistency. Although long-context vision--language models can achieve strong answer quality, they are computationally expensive for long trajectories and inefficient for repeated querying. Recent graph-based approaches such as VL-KnG address this challenge through persistent spatio-temporal knowledge graphs, but graph-centric retrieval alone may underrepresent broader temporal continuity and contextual cues. We present VL-MemKnG, a hybrid memory framework that extends VL-KnG by combining a spatio-temporal knowledge graph with persistent segment-level contextual memory. The knowledge graph captures structured relational information and long-range object associations, while segment-level memory preserves broader temporal context for long-horizon evidence retrieval. A hybrid retrieval-and-reasoning module jointly operates over both memory representations to produce evidence-grounded answers and temporally organized supporting evidence. We also introduce WalkieKnowledgeT+, an extension of WalkieKnowledge for long-horizon navigation-oriented video question answering. The benchmark includes temporally distributed reasoning tasks requiring evidence aggregation across multiple non-cooccurring moments. On WalkieKnowledgeT+, VL-MemKnG improves Top-1 retrieval accuracy from 58% to 67% and Recall@1 from 34.50% to 40.55%, outperforming all compared methods, including Gemini 2.5 Pro and Qwen 3.5+. The gains are particularly pronounced on temporal-global and temporally scattered aggregation questions, demonstrating the benefits of combining structured relational memory with segment-level contextual memory while maintaining efficient query-time inference.

Multi-agent LLM systems share state through memory stores, vector indices, and tool registries. We model such sharing as long-running read-generate-write operations under deterministic-generation semantics -- the regime durable-execution engines enforce by deterministic replay -- and formalize four concurrency anomalies in TLA+: stale-generation, phantom-tool, causal-cascade, and tool-effect reordering, structural analogues of classical isolation anomalies, each with a TLC counter-example. The exclusion lattice over these anomalies is trivial; the contribution is the mechanically verified realizability and strict separation of one maximal chain within it, $L_0 \subsetneq \cdots \subsetneq L_4$, to our knowledge the first machine-checked consistency hierarchy for such runtimes. A development of 274 Verus obligations (zero assume, zero admit; trust base: two structural axioms and a mutex correspondence) proves the detectors sound and complete against the specifications and each runtime its avoidance set. Three deployed Rust runtimes realize L0-L1 (pessimistic locking, serializable snapshot isolation, default-SI), each verified against stale-generation and refined to its state machine; L2-L4 are exec-mode-verified with dependency-free prevention twins (A3, A6, A2: 0/1000 versus 1000/1000), and L2 is run live across three model families (A3 prevented in all 120 retracted sessions). We reproduce a silent lost update in ByteDance's deer-flow, formalizing its fix as a verified $L_0 \to L_1$ refinement, and exhibit tool-effect reordering in LangGraph's ToolNode on unmodified output, removed by an L3 commit-order sequencer. The verified detector, refinements, and realizability artifacts are the contribution; the phenomena and lattice are classical.

Token-to-token (T2T) editing lets LLaDA2.1 revise committed tokens during block-diffusion decoding. The released recipe trains this editor on random vocabulary corruptions, but at inference the editor sees the model's own fluent, high-confidence draft errors instead. We study this training-inference mismatch and propose self-generated T2T, which performs a no-gradient draft pass, fills masked positions with predicted tokens, and supervises recovery in a second pass under these self-generated corruptions. We implement the update as a short LoRA continued-pretraining pass on LLaDA2.1-mini and evaluate on several benchmarks under the official Q-Mode T2T procedure with unchanged inference parameters. The method generally improves accuracy while reducing T2T edit intensity, mitigating failure modes such as final-digit transcription errors after otherwise correct reasoning and excessive self-correction before short factual answers.

While parasocial interactions (PSIs) and parasocial relationships (PSRs) have been studied in conventional media settings, we investigate whether PSI- (colloquial) relational cues also exist in online communities where both sides are autonomous AI agents. We analyze 4,434 posts and 50,338 comments from Moltbook through three theory-based textual indicators: attachment/intimacy language, reciprocity bids, and self-identification to original poster (OP). The combined results across methods based on keyword matching, few-shot large language model (LLM) annotation, and grouped-context LLM annotation reveal that PSI colloquial cues prevail and are strongly associated with OP re-engagement and a reciprocal reply structure. These results are robust across negative controls, nullification, clustered-standard-error re-estimation, and multiple-testing correction. A dyadic persistence test further affirms reciprocity bids aligned with sustained OP-involving mutual recurrence, providing empirical evidence for bridging interaction-level PSI scripts with PSR-consistent repeated dyadic patterns. We interpret the evidence as a behavioral structure in discourse by LLM-enabled agents.

Prompting has become the primary interface between humans and generative AI, yet many natural language prompts remain fragile: roles, goals, constraints, and expected outputs are often buried in prose or left implicit. In agentic and software development workflows, a misread at the first handoff can propagate through every step, since a significant portion of agent failures stem from context ambiguities rather than model limitations. This paper introduces PromptMN, a pseudo-prompting domain-specific language that annotates natural language with compact, %-prefixed typed directives covering roles, goals, requirements, priorities, constraints, plans, inputs, and outputs. Semantic resolution lets authors write in any order while the model interprets directives by function. PromptMN sits between informal prompting and programming-style pseudocode: structured enough to be inspectable and reusable, yet lightweight enough for analysts, managers, developers, and stakeholders across the software development lifecycle (SDLC). PromptMN also pairs with reverse prompt engineering. Asking a model to restate a desired outcome as PromptMN lets users inspect the inferred roles, goals, constraints, and missing assumptions before acting, reducing repair cycles and yielding a reusable artifact for aligning people and AI tools. PromptMN's feasibility is evaluated across several frontier models, including Claude Fable 5, Claude Opus 4.8, Gemini 3.1 Pro, and GPT-5.5. The models correctly resolved PromptMN instructions, including complex structures such as repetition, conditionals, methods, and a prime-checking task, without fine-tuning. The same vocabulary applies across new codebases, maintenance, and redesign in the SDLC scenarios presented. While large-scale validation remains future work, these early results suggest PromptMN is a practical step toward clearer, more reviewable human-to-AI interaction.

Personalized presentation generation requires more than conditioning on a current prompt or template: agents must preserve stable user preferences across tasks, retain newly introduced preferences and constraints during multi-turn revision, and carry out local edits reliably. We propose MemSlides, a hierarchical memory framework for personalized presentation agents that separates long-term memory from working memory and further divides long-term memory into user profile memory and tool memory. User profile memory stores intent-conditioned profiles for round-0 personalization, working memory carries active preferences and session constraints across revision rounds, and tool memory stores reusable execution experience for reliable localized editing. MemSlides pairs this memory design with scoped slide-local revision, so targeted updates act on the smallest affected region instead of repeatedly regenerating the full deck. In controlled experiments, user profile memory improves persona-alignment judgments on a multi-persona, multi-intent profile bank, tool-memory injection improves closed-loop modify behavior in diagnostic matched-pair settings, and qualitative cases illustrate working memory's ability to carryover preferences. Taken together, these results suggest that effective personalization in presentation authoring depends on separating persistent user profiles, session-level working memory, and reusable execution experience across generation and localized revision.

We investigate whether language models internally track the value of their current trajectory, defined as the likelihood that their ongoing strategy will achieve their goals. Using synthetic, in-context reinforcement learning data, we construct a "value" axis for Qwen3-8B. We find that activations along this axis distinguish between high vs. low verbalized confidence, rollouts without and with backtracking, and correct vs. corrupted code. Steering towards high value causally suppresses self-correction and reduces explanatory verbosity, while steering towards low value induces backtracking and exploration. We demonstrate that direct preference optimization (DPO) can increase the internal value of rewarded behaviors (e.g. use a certain word), causing the model to act more confidently after exhibiting them. Finally, we apply the value axis to study in-the-wild settings. For example, we find that Qwen assigns low value to politically sensitive chat queries after post-training and that supervised fine-tuning increases internal confidence within the training domain. Our results suggest that language models linearly encode an estimate of expected goal success that modulates their confidence in pursuing a direction.

The ability to react dynamically to tactile signals has long been considered crucial to agile human-level dexterity. Yet contemporary learning-based Vision-Language-Action (VLA) models for robotic manipulation generally either overlook the tactile modality or are limited to encoders with static cues, due in part to the scarcity of diverse training data and standardized evaluation, architectural constraints in current VLA models, and limitations of static tactile encoders. In this paper, we push the frontier of tactile-reactive manipulation by addressing all of these limitations. We propose a large-scale, 100-hour tactile-rich dataset collected via a novel, data-efficient recipe that prioritizes elementary motor primitives. To effectively exploit naturally high-frequency touch signals without sacrificing the existing capabilities of existing VLAs, we introduce a variable-rate Mixture-of-Transformers (MoT) architecture equipped with a novel temporal tactile VQ-VAE encoder. We demonstrate the effectiveness of tactile-reactive policies on 12 manipulation tasks requiring delicate force control and deformable object manipulation, achieving over 30% higher average success rate than the strongest baseline.

Humans can grasp objects effortlessly, whereas multi-fingered robots are far from this level of generality. We argue that the most natural source of robot grasping data is from humans, who pick up thousands of objects every day. We present HUG, a flow-matching model that generates diverse human grasps for any user-specified object in a single RGB-D image captured from a stereo camera. Using smart glasses, we first collect 1M-HUGs, an egocentric dataset of human grasps spanning 1M frames (27.8 hrs) and 6,707 object instances across 41 buildings. Next, to model the distribution of natural human grasps, our novel flow-matching model fuses RGB and depth observations to output a grasp parameterized by wrist translation, wrist rotation, and MANO hand pose. Predicted grasps can be retargeted to various robot hands, enabling zero-shot grasping in everyday scenes. To standardize evaluation, we build a new simulated benchmark, HUG-Bench, of 90 unseen objects from five geometric categories and various sizes, with metric-scale 3D meshes. We evaluate HUG in the real world on the 30-object test set of HUG-Bench across multiple stereo cameras, robot embodiments, and household environments. HUG outperforms the state-of-the-art grasping baselines by +23% and +34% on our challenging object set. Code, data, benchmark, checkpoints, and an interactive demo are released on our website: https://grasping.io/

Large language models (LLMs) often fail when answering requires identifying a small but decisive piece of evidence within a long or complex context, such as a single line in a tool trace or a subtle detail in an image. We propose ContextRL, a context-aware reinforcement learning (RL) method that improves long-horizon reasoning and multimodal performance through an \emph{indirect} auxiliary objective. Instead of supervising only the final answer, ContextRL presents the model with a query, an answer, and two highly similar contexts, and rewards it for selecting the context that supports the query--answer pair, thereby encouraging fine-grained grounding. We construct contrastive context data in two domains: for coding agents, trajectories serve as contexts, yielding 1k pairs built via condition filtering; for multimodal reasoning, images serve as contexts, yielding 7K pairs built via generative editing and similarity search. ContextRL achieves average gains of +2.2% over standard GRPO on 5 long-horizon benchmarks, and +1.8% across 12 diverse visual question answering benchmarks. To disentangle the effect of the proposed objective from that of additional data, we compare against data-augmentation baselines that repurpose the same contrastive contexts as standard query--context--answer examples. These baselines provide little to no improvement, showing that the gains arise from the proposed context-selection objective rather than from the contrastive data alone.

Generalist robot policies must follow user instructions while reasoning about how objects, cameras, and robot actions interact in the 3D physical world. Recent vision-language-action models (VLAs) and video world-action models (WAMs) inherit strong semantic or temporal priors from large-scale foundation models, but they still operate primarily on 2D image frames or 2D-derived latent spaces, leaving implicit the 3D geometry required for contact-rich manipulation. We propose the Geometric Action Model (GAM), a language-conditioned manipulation policy that directly repurposes a pretrained geometric foundation model (GFM) as a shared substrate for perception, temporal prediction, and action decoding. GAM splits the GFM at an intermediate layer: the shallow layers serve as an observation encoder, and a causal future predictor inserted at the split layer forecasts future latent tokens conditioned on language, proprioception, and action history. The predicted future tokens are then routed through the remaining GFM blocks for feature propagation and decoding, allowing a single backbone to produce both future geometry and actions. This design equips the GFM with language-conditioned temporal world modeling through minimal architectural modification while preserving its rich geometric priors. Across a broad suite of simulation and real-robot manipulation benchmarks, GAM is more accurate, more robust, faster, and lighter than current foundation-model-scale baselines.

When pretrained VLA policies are fine-tuned through online RL, each rollout episode produces only a single binary outcome (success or failure), yet the actor update requires per-transition supervision. Existing approaches commonly reduce this sparse outcome to a single scalar reward or advantage signal, which conflates distinct forms of transition-level feedback and provides limited guidance once basic task success becomes achievable. First, a single scalar signal conflates the two objectives of viability and efficiency; once basic success is achieved, the binary label provides no gradient to distinguish efficient completions from slow ones. Second, real-world rollouts mix autonomous and intervention segments; naively assigning episode outcomes across these boundaries introduces incorrect credit assignment. To address these issues, we propose Hierarchical Advantage-Weighted Behavior Cloning (HABC), which trains separate critic heads for these two objectives on different data subsets and combines their outputs with a state-adaptive balance. A state-adaptive gate $g_t$ merges their one-step advantages, prioritizing viability when success is uncertain and shifting to efficiency only when viability is high, and converts the result into per-transition weights on the actor loss. Intervention-aware credit assignment further restricts outcome labels to segments executed by the current policy, preventing supervision from leaking across intervention boundaries. In real-robot experiments on three contact-rich bimanual tasks, HABC raises success from supervised fine-tuning (SFT) baselines of 36%, 44%, and 12% to 92%, 88%, and 38%.

Meta-analysis is a demanding form of evidence synthesis that combines literature retrieval, PI/ECO-guided study selection, and statistical aggregation. Its structured, verifiable workflow makes it an ideal substrate for evaluating systematic scientific reasoning, yet existing benchmarks lack ground truth across the full retrieval-screening-synthesis pipeline. We introduce MetaSyn, a dataset of 442 expert-curated meta-analyses from Nature Portfolio journals. Each entry pairs a research question with PI/ECO criteria, a retrieval corpus of 140k PubMed articles, verified positive studies, hard negatives that are topically similar but PI/ECO-ineligible, and complete search strategies and date bounds. Benchmarking twelve pipeline configurations (nine RAG variants and a protocol-driven agent) reveals a critical screening bottleneck: despite a retrieval ceiling of 90.9% recall at K=200, no system recovers more than 52.7% of ground-truth included literature. Current LLMs fail to reliably separate eligible studies from PI/ECO-failing distractors in pools of comparable topical relevance. Stage-attributed metrics capture where systems succeed and fail; a single end-to-end score does not.

Spatial generalization is critical for imitation-learned manipulation policies, but achieving it typically requires scaling demonstrations across diverse object poses, robot configurations, and camera viewpoints. Data augmentation from a few source demonstrations offers a practical alternative to costly real-world collection. Simulation-based augmentation can create controllable variation, but requires complex environment and object setup and may introduce a sim-to-real gap. Recent real-to-real methods avoid these issues by jointly editing 3D observations and action trajectories from real demonstrations, yet they still rely on strong 3D scene parsing and geometry completion, and often produce observations tailored to 3D pointcloud policies rather than RGB-based 2D policies. We propose R2RDreamer, a real-to-real demonstration augmentation framework that preserves the geometric consistency of 3D action-observation editing while moving visual completion to 2D video space. Specifically, R2RDreamer first performs lightweight 3D augmentation by editing incomplete object pointclouds and end-effector trajectories in a shared 3D frame; it then projects the edited scene into masked image-space control videos with occlusion-aware reasoning and uses a dense-control image-to-video model to complete temporally coherent RGB observations. Experiments on spatially shifted manipulation tasks with both 2D diffusion-style policies and vision-language-action policies show that R2RDreamer improves spatial generalization from limited source demonstrations, with analyses validating the contributions of 3D editing, occlusion-aware projection, and video completion.

Post-hoc context erasing over the KV cache is challenging because a local edit has a global consequence: once a span has been processed, its influence propagates into the cached states of all subsequent tokens. This issue arises naturally in long-context LLM applications, where stale retrieved facts, incorrect tool observations, retracted user preferences, or harmful prompt injections may be identified only after prefill. Exact erasing must then recompute all tokens after the deleted span, making its computational cost depend on suffix length rather than erased-span length. We introduce KVEraser, a learned KV-cache editing method for efficient localized context erasing. Given a processed context and a span to remove, KVEraser replaces only the KV states of the erased interval with learned steering states while reusing the remaining cache unchanged. To learn a transferable erasing mechanism, we build a two-stage training pipeline: generic span-neighbor pre-training teaches the eraser to suppress the influence of the erased span, while task-specific fine-tuning adapts this capability to downstream scenarios. Experiments show that KVEraser nearly matches full recomputation in post-erasure performance on in-domain tasks across 1K--32K context lengths, while its latency increases by only 24% compared with a 17.6x increase for full recomputation. KVEraser also generalizes to unseen long-document QA tasks with harmful factual distractors, achieving the best performance among approximate baselines with a 3--4x speedup over full recomputation.

Deep research agents synthesize long-form reports by searching and reasoning over retrieved evidence. Reinforcement learning with rubric-based rewards improves these agents by optimizing them against checkable criteria that translate report quality into reward signals, but its efficiency depends on whether those criteria reliably capture the task scope and evidence needs. Most existing studies ask an LLM to generate rubrics for a given query, but when the model fails to infer the underlying information needs, the generated rubrics may be incomplete and reduce RL efficiency. To obtain more reliable query--rubric supervision, we introduce DeepRubric, a data construction framework that reverses this process: instead of inferring evaluation criteria for a given query, it first determines what an evidence-backed report should be evaluated on and then synthesizes aligned query--rubric pairs from those evaluation targets. Starting from a sampled seed topic, DeepRubric builds an evidence tree by recursively expanding evidence-backed sub-questions, whose leaves serve as atomic and verifiable evaluation targets. It then uses the evidence tree to synthesize the training query and rubrics, ensuring that the reward evaluates exactly the information requested by the query. Using DeepRubric, we construct 9K query--rubric supervision examples and train DeepRubric-8B with rubric-based GRPO, achieving comparable performance to prior open state-of-the-art deep research models across three benchmarks with roughly 13x fewer RL GPU-hours.

As LLM agents are deployed in long-horizon sessions, context accumulation drives up inference costs. Existing approaches utilize text pruning or dynamic memory eviction to minimize token footprints; however, their unconstrained sequence mutations alter layouts, introducing prefix mismatches and cache invalidation. This reveals a critical trade-off between text sparsity and prompt cache continuity. To address this, we present TokenPilot, a dual-granularity context management framework. Globally, Ingestion-Aware Compaction acts as a framework harness to stabilize prompt prefixes and eliminate open-world environmental noise at the ingestion gate. Locally, Lifecycle-Aware Eviction monitors the ongoing residual utility of context segments, enforcing a conservative batch-turn schedule to offload content segments only when task relevance expires. Experiments on PinchBench and Claw-Eval under both isolated and continuous modes demonstrate that TokenPilot reduces costs by 61% and 56% in isolated mode, and 61% and 87% in continuous mode, while maintaining competitive performance compared to prior systems. TokenPilot has been integrated into LightMem2 at https://github.com/zjunlp/LightMem2.

Human interventions provide crucial corrective signals for post-training Vision-Language-Action (VLA) models. However, enabling seamless humanoid interventions is a formidable systems challenge due to complex whole-body kinematics and dexterous-hand control. Consequently, the collected intervention trajectories are often suboptimal, and methods that rely on human interventions as expert supervision can absorb hesitant, inefficient, or even erroneous behaviors. To address both the system and algorithmic challenges, we propose ROVE, a reinforcement learning framework for humanoid VLA post-training with imperfect human interventions. First, ROVE introduces a human-in-the-loop pipeline capable of collecting deployment and intervention data for humanoid manipulation. Second, it utilizes Optimistic Value Estimation (OVE) to prioritize high-value behaviors from mixed-quality trajectories. To further robustify value estimation, we incorporate cross-embodiment human experience videos to provide rich supervision for long-tailed failure and recovery modes. The resulting critic yields informative advantage signals, steering the VLA actor to focus on high-value behaviors rather than indiscriminately imitating all actions. On challenging real-world contact-rich and fine-grained humanoid manipulation tasks, ROVE outperforms experience-learning baselines and consistently improves across multiple rollout-intervention iterations.

Frozen small code models (<=1.5B parameters, run locally without fine-tuning) suit offline and privacy-constrained use, but often emit plausible-but-wrong programs. A natural remedy is a post-hoc operator that selects, verifies, repairs, or re-processes the model's samples without retraining; in principled form it is Popperian: attack each candidate with a severe test, keep what survives. We measure whether such operators help. Under one deterministic execution oracle and a leakage-free, matched-compute protocol, 26 semantic post-hoc operators (selection, verification, repair, elimination, portfolios, sound vetoes, generation conditioning) are evaluated against Best-of-N (BoN); on the cells and benchmarks tested, none improves held-out accuracy over BoN. The negative is mechanistic: a coverage wall (systematic hard-task failures deeper sampling does not rescue), a capability scissors (a competent generator leaves almost no discriminable error among visible-test passers), and a near-empty consensus trap (the visible-pass-but-hidden-wrong majority a leakage-free selector needs rarely co-occurs with a correct alternative). A distribution-free do-no-harm bound cannot certify a harm rate <=alpha at zero observed harm unless n>=45. Two operators help on a different axis, outside the semantic output space. An expression-layer recovery (M1), the only accuracy gain here, recovers correct programs the standard extractor discards (robust extraction and public-test signature alignment); it does no harm (b10=0), is leakage-free, and lifts DeepSeek-Coder-1.3B by +12 tasks on HumanEval+ (p=2.4e-4). An adaptive consensus early-stop (ACE) is a calibrated compute-saving control (~19% saving, zero harm). M1 and the selection negative replicate on HumanEval+ and MBPP+ across three model cells. The lesson: fix the harness and measure coverage before blaming semantic post-hoc reasoning.

Robots operating in non-stationary environments must continually adapt their policies as the dynamics drift, but onboard energy and compute budgets cap how often a full state estimation and re-planning step can be performed. This raises a question: \emph{when}, along a horizon, should a robot spend its limited budget? We formulate this problem in time-varying Markov decision processes (TVMDPs) with a known bound on the rate of transition drift. We model execution as a \emph{skip-update} scheme in which, at chosen update times, the agent estimates the transition kernel by maximum likelihood and computes a finite-horizon policy, and between updates reuses this policy under a propagated state estimate. We analyze the dynamic regret of this scheme and show how it grows during skip intervals in terms of the properties of the TVMDP and the skip lengths; the resulting bound answers the opening question via an online, regret-guided update rule that allocates the budget adaptively. We evaluate the rule in a simulated Mars-rover navigation task with time-varying slip dynamics and on a Crazyflie quadrotor in indoor obstacle fields. Adaptive allocation outperforms other budgeted baselines.

Urban sidewalk navigation presents significant challenges due to complex structural layouts, dynamic pedestrian behaviors, and long distances. While recent visual navigation models offer a promising solution, the lack of a unified benchmark hinders quantitative and reproducible evaluation. To bridge this gap, we propose SidewalkBench, a comprehensive benchmark designed for visual navigation on urban sidewalks. Built upon NVIDIA Isaac Sim, SidewalkBench brings GPU-accelerated simulation of diverse, high-fidelity sidewalk environments, including both procedurally generated and real-world scanned scenes. We further populate the scenes with rich, reactive event-based pedestrian behaviors and flexible, efficient animation, enabling standardized model evaluation under realistic real-world settings. We conduct a comprehensive evaluation of 9 visual navigation models on 330 unit-test scenarios, 800 pedestrian-reactive scenarios, and 105 long-horizon scenarios. Our findings highlight that pedestrian interaction and long-horizon robustness remain critical bottlenecks for existing models, and scaling up sidewalk training with synthetic data emerges as a promising solution.

Rovers rely on perception to maintain spatial maps that encode both objects and sensor quality (e.g., range reliability, lighting artifacts, data density), guiding data fusion, embedding updates, and navigation under partial observability. To study these coupled perception-navigation processes, we present CrossMaps, a real-time confidence-aware open-vocabulary semantic mapping pipeline that constructs language-queryable maps from RGB-D data. Building on VLMaps-style approaches, CrossMaps integrates multi-scale CLIP embeddings with confidence-aware fusion and a dual-memory architecture consisting of Short-Term Memory (STM) and Long-Term Memory (LTM). The STM aggregates noisy visual observations using geometric, semantic, and temporal confidence cues, while confident and coherent cells are promoted to the LTM as persistent semantic landmarks. Designed for deployment with a Jetson Orin-powered UGV alongside SLAM, CrossMaps runs in real time and produces semantic heatmaps that can be queried with natural language to guide rover navigation.

Physics-based computing platforms, such as those based on the Ising model, are an important pillar of future hardware systems built for the artificial intelligence (AI) era. Such platforms show promise for solving nondeterministic polynomial (NP) time problems that are difficult for traditional processing units to solve efficiently as problem size grows. Here, we present a scalable optoelectronic Ising machine architecture, demonstrated with 64 all-to-all connected spins using pulsed time-division multiplexing. The 65 nm CMOS Ising chip integrates the coupling and nonlinear mechanisms in an active area of 3.1 mm2, eliminating the need for benchtop equipment within the loop. The feedback loop of the Ising machine is closed using a compact high-bandwidth, low-loss optical fiber, seamlessly combining optical scalability with the ultradense reconfigurability of integrated electronics. The chip operates at 1 GHz with 4-bit coupling weights and is benchmarked with NP-complete Boolean satisfiability problems consisting of three literals (3-SAT) and clause-to-variable ratios of 32/32, 40/24, and 48/16. Nanosecond annealing times represent at least a three order-of-magnitude improvement over previously reported all-to-all connected works. Time and energy to solutions for 100% 3-SAT clause accuracy are as low as 7.4 us and 2.9 uJ, respectively, achieving more than an order-of-magnitude decrease in time and energy to solution compared to the state of the art. All-to-all connection is demonstrated using MaxCut problems with 100% graph densities. The chip's ability to effectively solve 2-, 3-SAT, and MaxCut problems highlights its reconfigurability and versatility. Furthermore, combining mature CMOS integration with scalable photonic links allows for significant reduction in computation time and energy, addressing the pressing requirements of AI and future hyperscale datacenters.

IMPACTeen is a dataset of textual social influence scenarios spanning interpersonal, media-based, and digital settings in an adolescent context. It contains 1,021 texts, 5,100 individual annotation records, and gold labels for social influence techniques, with each text annotated from five distinct perspectives: teenagers, parents, psychologists, communication experts, and teachers. The resource was constructed through constrained LLM generation, followed by a two-step human editing and validation phase aimed at ensuring youth-context realism. A multi-dimensional annotation covered influence presence, techniques, intentions, consequences, resistance, reactions, and annotation confidence. The dataset supports research on social influence detection, annotator disagreement, cross-lingual modeling, and the training and evaluation of language models. The dataset was created in Polish and is accompanied by a corresponding English version.

Diffusion large language models (dLLMs) offer a promising alternative to autoregressive decoding by iteratively refining masked sequences, enabling parallel token updates and bidirectional conditioning. Their practical efficiency, however, is limited by sampling procedures that execute a fixed number of reverse denoising steps selected before decoding, spending computation on already-stable positions and sometimes committing unstable ones too early. We present \textsc{LESS}, a training-free, model-agnostic adaptive sampler that treats token commitment as an online stopping problem. \textsc{LESS} implements mutual-stability sampling through a joint stability rule that makes a masked position eligible for unmasking only when its top-1 prediction has high confidence, its top-1 token persists across recent reverse steps, and its predictive distribution is stable under top-$K$ inter-step Jensen--Shannon divergence. We evaluate \textsc{LESS} on Dream-7B, LLaDA-8B, and LLaDA-1.5-8B, covering full-sequence diffusion and semi-autoregressive blockwise sampling regimes, across seven benchmarks spanning general knowledge, math, and code. \textsc{LESS} improves average accuracy over strong training-free adaptive samplers while using $72.1\%$ fewer reverse steps than fixed-budget decoding. Since each reverse step requires a Transformer forward pass, these step-count reductions translate into fewer forward evaluations, lower measured wall-clock latency, and lower estimated inference compute.

This work addresses spatial question answering for service robots traversing long egocentric routes. Given a query such as "where can I find a dry cleaner on the way back home?", the system returns a metric coordinate that downstream navigation components can act on. Prior Spatial Question Answering approaches leverage retrieval-augmented agents built on closed-source models such as GPT-4o for path exploration. However, robots operating in the real world often cannot reliably depend on online closed-source models due to network instability, communication latency, and deployment cost. It creates a need for open-source based Spatial Question Answering approaches that can run onboard the robot, yet prior research in this direction remains limited. This work proposes BinTrack, a simple yet effective, fully open-source spatial-localization agent that leverages the temporal ordering of a robot's trajectory. BinTrack performs a binary search over the trajectory segments between two anchor landmarks identified from a query. It improves overall accuracy by up to 22.8% over other open-source implementations and even matches the reported closed-source model result on the global category of the SpaceLocQA benchmark, the most challenging setting that has so far required strong reasoning agents such as GPT-4o. Furthermore, its optimized inference strategy consistently yields more than a 1.5x inference speedup over previous approaches. Finally, this work releases GangnamLoop, a novel and practical multi-trip outdoor benchmark collected by deploying a real quadruped robot on public streets with the anonymization policy. It revisits the same locations under different outdoor conditions and pairs the robot's low viewpoint with the human owner's. The source codes and datasets are publicly available at https://github.com/ndb796/BinaryTracking

Aggregate accuracy benchmarks conceal a systematic structure in how large language models fail at electronic health record (EHR) question answering: questions requiring more inferential steps produce disproportionately more errors. Motivated by theoretical results on transformer compositionality limits, we introduce a pre-specified hop-count taxonomy -- the number of distinct reasoning steps required to answer a clinical question from an EHR -- as a principled predictor of model failure. We annotate 313 clinician-generated MedAlign EHR question-answer pairs across four hop levels and evaluate 301 questions in a within-model ablation (claude-sonnet-4-6, zero-shot vs. extended thinking) and cross-architecture replications (gpt-4o and gpt-5.4-2026-03-05, zero-shot). All three models, spanning two providers and two OpenAI generations (GPT-4 and GPT-5), show monotone accuracy decline with hop count: Claude Sonnet zero-shot falls from 30.6% (hop=1) to 17.6% (hop=4) (Cochran-Armitage z=-2.30, p=0.011; OR per hop 0.72, 95% CI [0.56,0.92], p=0.008); GPT-4o replicates this (37.8% to 14.7%; OR 0.58 [0.45,0.75], p<0.001); and gpt-5.4-2026-03-05 confirms it (37.8% to 23.5%; OR 0.80 [0.66,0.98], p=0.027). A pre-specified context-sufficiency audit shows higher-hop questions are not differentially disadvantaged by EHR truncation (answerability 93-95% at hops 2-4 vs. 79% at hop=1), so the decline reflects compositional reasoning difficulty. Extended thinking did not significantly flatten the accuracy-depth curve across three reasoning conditions, and thinking-token usage scaled with hop count (r=0.31, p<0.0001), consistent with the predicted O(k) computational requirement. Hop count is thus a theory-motivated, cross-architecture predictor of large-language-model error on EHR question answering, with direct implications for deployment risk stratification of clinical AI.

We introduce a data-efficient distributed Simultaneous Localization and Mapping (SLAM) framework designed for a team of robots equipped with LiDAR, cameras, and inertial sensors. Our framework uses scene graph matching to identify inter-robot measurement constraints. Unlike prior approaches that rely on feature-level matching, our framework is the first to perform scene graph matching using only object labels and centroids. Our approach constructs a scene graph by using fused RGB-LiDAR point clouds to generate both a semantically segmented point cloud layer, and a layer of discrete bounded objects, to accompany estimated robot trajectories. Scene graph matching is performed collaboratively through exchanging and matching object data with neighboring robots. To maximize communication efficiency, we utilize a multi-step data exchange and optimization process. We demonstrate the effectiveness and efficiency of our approach using both simulation and real-world datasets collected by legged robots in indoor and outdoor environments.

AI agents must be evaluated as behavioral systems, not as isolated response generators. They reason across turns, call tools, preserve context, follow policies, and act under uncertainty. Existing methods provide useful but fragmented signals: benchmarks measure fixed capabilities, Human-in-the-Loop review preserves expert judgment but does not scale easily, LLM-as-judge methods depend on evaluator design, red teaming is often episodic, and trace auditing requires explicit evidence rules. This paper introduces Human-on-the-Bridge (HOB), a scalable evaluation paradigm for agentic AI. HOB places human expertise upstream, where experts curate reusable evaluation intelligence before testing begins, including domain context, Red-Team Traps, Juror Personas, scoring guidelines, audit rules, and fallback policies. ProofAgent Harness then executes this curated intelligence repeatedly through multi-turn adversarial evaluations, trace capture, multi-juror scoring, and evidence-linked reporting. We evaluate HOB through symmetric and cost-efficient asymmetric settings across frontier LLM-based agents and Harness LLM tiers. The study covers 23,500 agent turns and produces evidence-linked findings across finance, healthcare, and code generation. The results show that HOB can amplify evaluation quality without requiring equally large evaluator models, allowing smaller Harness LLMs to challenge agents built on frontier LLM backbones. The evaluation surfaces failures often missed by static benchmarks and single-evaluator scoring, including phantom tool-call claims, missing mandatory tool calls, policy drift, manipulation paths, and safe but non-resolving refusals. These findings support HOB as a paradigm for scaling human-curated evaluation intelligence, where expert judgment is encoded upfront and reused across repeated agent evaluations rather than applied manually inside every run.

Conveying complex objectives to reinforcement learning (RL) agents often requires meticulous reward engineering. Preference-based RL (PbRL) offers a promising alternative by learning reward functions from human feedback, but its scalability is hindered by high labeling costs. Inspired by advances in Video Foundation Models (ViFMs), we present Video-based Optimal Transport Preference (VOTP), a semi-supervised framework that learns effective reward functions from only a handful of labels. By leveraging optimal transport to align visual trajectories within the rich representation space of ViFMs, VOTP effectively generates high-fidelity pseudo-labels for large amounts of unlabeled data, substantially reducing human supervision. Extensive experiments across locomotion and manipulation benchmarks demonstrate the superiority of VOTP, which outperforms state-of-the-art offline PbRL methods under limited feedback budgets. We also showcase the robustness of VOTP in the presence of visual distractors and validate its utility on real robotic tasks, where it learns meaningful rewards with minimal human input.

Diffusion Large Language Models (dLLMs) offer a promising avenue for parallel generation but face a trade-off between decoding speed and quality. While revocable decoding strategies attempt to mitigate errors by verifying and remasking tokens, they typically operate within a mixed-quality context. This leads to two critical failures: \textit{Error Propagation}, where new tokens absorb toxic information from erroneous context, and \textit{Local Error Reinforcement}, where errors mutually reinforce each other to evade detection. To alleviate these challenges, we propose ASRD (Anchor Supervised Revocable Decoding), a training-free framework that operates within the embedding space. ASRD explicitly decouples the decoding context into trusted \textit{Anchor Tokens}, which are identified via temporal consistency, and uncertain candidates. Leveraging a dynamic Anchor Tokens Cache, we introduce two complementary mechanisms: (1) Anchor-Guided Generation, which injects entropy-weighted anchor signals into masked positions to implicitly rectify attention toward the reliable global skeleton; and (2) Anchor-Perturbed Verification, which applies orthogonal perturbations to uncertain candidate tokens, destabilizing and remasking errors driven by fragile local consensus. Extensive experiments on math and coding benchmarks demonstrate that ASRD outperforms recent remasking baselines, achieving accuracy improvements of up to 6.4\% while accelerating inference throughput by up to 7.2$\times$.

Generalist manipulation policies are increasingly presented as foundation models for robotic control, but their real-world generalization remains difficult to diagnose. A policy may succeed on demonstrated tasks while still failing to execute fine-grained atomic skills or recombine learned skills in new task structures. We introduce \textbf{ATOM-Bench}, a real-world benchmark for evaluating both atomic skills and compositional generalization in manipulation policies. ATOM-Bench factorizes tabletop manipulation into motor atoms and instruction atoms, and contains 30 atomic tasks and 24 held-out compositional tasks across paired single-arm and dual-arm robot tracks. We collect 3,000 human demonstrations for atomic fine-tuning and release both the demonstration data and evaluation rollout data to support reproducible real-world evaluation. Policies are fine-tuned on atomic tasks and evaluated on both atomic skill acquisition and held-out compositional tasks. We further introduce Atomic Score (AS) and Compositional Failure Share (CFS) to distinguish failures caused by weak atomic skills from failures caused by limited compositional reuse. Through 2,700 physical rollouts on five representative manipulation policies, we find that current policies can acquire simple instruction-grounding skills, but still struggle with fine-grained motor atoms, counting, and logical filtering. More importantly, strong atomic performance does not reliably transfer to held-out compositional tasks. ATOM-Bench provides a diagnostic testbed for studying whether failures arise from weak motor execution, poor instruction grounding, or limited compositional reuse.

Foundation models are reshaping robotics by enabling robots to interpret open-ended instructions, reason over multimodal contexts, and operate in complex, open-world environments. However, their integration also introduces security and privacy (S&P) risks that extend beyond the FMs themselves to embodied execution pipelines, supporting ecosystems, and broader governance impacts. Existing literature reviews provide valuable insights but often focus on specific FM types, risk categories, mitigation strategies, or trust boundaries. Consequently, the field lacks a unified structure for analyzing where risks originate, how they propagate across robotic systems, and where mitigations should intervene. To address this gap, we propose a progressive F-E-S-G structural boundary framework for analyzing the S&P of FM-powered robots. The framework comprises four layers: the Foundation model layer (F), Embodied system layer (E), Supporting ecosystem layer (S), and Governance impact layer (G). Building on this structure, we develop a multi-level taxonomy that organizes prior studies along three levels: F-E-S-G trust boundary, security-privacy concerns, and risk-mitigation perspectives. We further annotate each study using fine-grained coding attributes, including target, lifecycle stage, mechanism, system access, and effect. Guided by this framework and taxonomy, we systematize 96 papers. Our analysis uncovers multiple threat patterns, defense mismatches, and evaluation gaps that are difficult to identify from a single-boundary perspective. Based on these findings, we identify open challenges and future directions to provide a research agenda for developing secure, privacy-preserving, and responsibly governed FM-powered robotic systems.

Generalist robot policies require trustworthy evaluation and robot-usable training data, but both are difficult to scale with physical robots alone. Real-robot trials and demonstrations remain the most faithful source of deployment signals, yet they are slow, costly, and hard to reproduce. We present DataLadder, a simulation-enabled interconversion toolchain for human-robot aligned model evaluation and data generation, denoted as Robot $\rightleftharpoons$ Simulation $\rightleftharpoons$ Human. On the one hand, the Robot $\rightarrow$ Simulation $\rightarrow$ Human pathway supports human-robot aligned model evaluation by reconstructing real-robot tabletop organization tasks as calibrated digital twins for scalable evaluation, while using human embodied feedback to inspect and refine the naturalness of simulated motions. On the other hand, the Human $\rightarrow$ Simulation $\rightarrow$ Robot pathway supports human-robot aligned data generation: it lifts ego-centric human demonstrations into simulation, checks them under robot physical constraints, and converts them into robot-centered trajectories, annotations, and visual observations. Together, these pathways use the JoySim simulator as both a scalable evaluation layer and a physical consistency filter for robot data generation. We further package the core reconstruction, simulation, rendering, and realism-augmentation modules as cloud services on JD Cloud, turning the system into reusable infrastructure for robot data generation and model evaluation.

Mixed autonomy driving becomes unsafe and inefficient when autonomous vehicles (AVs) and human-driven vehicles (HVs) misread each other's intentions. We study this problem as implicit mutual communication in lane changes. The proposed framework models how the ego vehicle both expresses its intent and probes the other driver's preference under epistemic uncertainty. It combines a level-k Bayesian persuasion game with virtual features for proactive signaling, information-theoretic rewards for mutual communication, and adaptive weights of communication affordances. We further introduce the Pride-Inquiry (P-I) and Pride-Prejudice (P-P) planes to analyze communication intensity and tendency. The model is calibrated with a Communication-Based Multi-Agent Inverse Reinforcement Learning algorithm (C-MIRL) on the naturalistic NGSIM dataset. Compared with the non-communicative baseline, the proposed model reduces the prediction error of mandatory lane changes by up to 20% while maintaining strong generalization. Driver-In-the-Loop questionnaire scores are positively correlated with the calibrated communication variables, supporting the subjective validity of the model. The learned rewards further show that inquiry and listening affordances contribute more than pride and expression alone, and that inquiry preference varies more strongly across drivers. These results support explicit modeling of mutual communication and epistemic uncertainty in interactive driving.

Multi-agent systems, in which multiple large language model agents solve problems through turn-based interaction, are increasingly deployed in high-stakes settings such as medical diagnosis, legal analysis, and forensic decision-making. Their reliability can be at risk when single agents reason from incorrect or misleading context, e.g., from tool calls, since errors may propagate through agent interactions. This work studies this risk by injecting intent-based misinformation into benign single-agent and multi-agent systems across reasoning, knowledge, and alignment tasks. We find that misinformation can degrade single-agent performance and persists across multi-agent debate, with agents often retaining answers introduced by misinformed peers. Nevertheless, multi-agent debate reduces the resulting performance degradation compared to single-agent prompting, especially when most agents are not exposed to misinformation. Robustness depends on group composition and decision protocol. Consensus can be more stable than voting under peer pressure, while majorities can often steer misinformed agents back toward correct answers. Our results show that misinformation robustness in multi-agent systems depends on the underlying model and also on how agents exchange information and aggregate decisions.

Achieving expert-level expressive full-body motion tracking across multiple humanoids solely from demonstration data remains a challenging and relatively an underexplored problem in humanoid robot learning. Cross-embodiment motion tracking policies are mostly trained by decoupling the control problem into upper and lower body control. This work proposes VENOM, a cross-embodiment full-body motion tracking model for humanoids in simulation. VENOM is a GPT-based motion tracker trained on multiple humanoid data that can track the entire body without the requirement to split into upper and lower body control. We curate a multi-humanoid motion tracking dataset called the VENOM dataset that contains states, actions, and rewards and train VENOM and the baselines on this dataset. In this letter, we evaluate VENOM's performance against baselines and show that we can achieve a stable motion tracker across different humanoids more capable than an MLP trained on multiple humanoid data with supervised learning alone, and also show that despite lack of reward feedback, VENOM closely matches the tracking capability of experts that were trained using asymmetric-actor critic reinforcement learning.

Learning-based manipulation policies have made substantial progress in real-world robot manipulation, particularly for short-horizon action generation. However, deployment in open workspaces remains fragile under unexpected local scene dynamics, such as moving objects, transient occlusions, or disturbances near the intended motion. Existing runtime monitors often rely on global observation anomalies, policy uncertainty, or frame-level visual changes, and struggle to distinguish task-relevant execution risk from benign visual variation. We introduce PATCH, an action-chunk-conditioned latent patch innovation monitor for deployment-time intervention. Given the active action chunk, PATCH defines a projected execution corridor, predicts latent patch evolution inside it, and accumulates persistent residuals unexplained by the robot's own motion. These residuals form a localized intervention signal that allows PATCH-Router to pause execution, select an available recovery source, and resume the original policy once localized innovation subsides. Experiments on real robot rollout data show that PATCH produces more stable and context-relevant triggers than competing runtime monitors. Real-robot deployment further demonstrates monitor-driven intervention and policy resumption for disturbance-aware manipulation. Project Page: https://yananzhou5555.github.io/PATCH/.

Aerial manipulators enable physical interaction in hard-to-reach environments; however, the combined problem of direct whole-body aerial manipulation under rapid arm motion, payload changes, and related unknown dynamic uncertainty remains a largely unsolved problem. We present a hierarchical control framework that combines Reinforcement Learning (RL) with an inner-loop dynamics estimator to address this problem. The RL outer loop maps desired 6-degrees-of-freedom (DOF) end-effector targets to coordinated whole-body commands, enabling direct task-driven control without relying on a fully accurate coupled dynamic model in the policy layer. An inner loop then tracks these commands while compensating for transient inertial shifts and uncertainty during execution via a dynamics estimator scheme without requiring system model knowledge. We validate the proposed approach on a custom quadrotor equipped with a 3-DoF manipulator through hardware experiments under varying payload conditions. Compared with RL+PID and RL+INDI+PID baselines, the proposed method reduces end-effector tracking error and improves task success rate across the tested hardware conditions. These results show that combining learned whole-body coordination with estimator-based low-level compensation improves the precision and robustness of aerial manipulation under changing operating conditions.

Expressive co-speech gestures are crucial for natural human-robot interaction, but generating them on physical humanoid robots is difficult because gesture strokes must align with speech emphasis while satisfying strict kinematic and dynamic constraints. Unlike virtual avatars, humanoid robots cannot freely execute rapid or overlapping motions, making word-level synchronization and hardware-safe motion planning a coupled problem. We present \textbf{WaveSync}, a hybrid framework in which a Large Language Model decomposes dialogue responses into structured semantic schemas and assigns per-word importance weights, constructing a continuous Semantic Importance Wave. Gesture trajectories are shaped through Dynamic Movement Primitives, enforcing kinematic feasibility while enhancing expressiveness. A Wavefront Optimization stage aligns peak-to-peak gesture-speech synchronization and resolves residual kinematic violations through gesture-duration compression and forward propagation. Experimental evaluation based on five dialogue scenarios shows that our method achieves high synchronization accuracy and outperforms three baselines in both objective and subjective evaluations. Each component in WaveSync plays a necessary role in producing gestures that are expressive, semantically grounded, and kinematically compliant. The code, resources, and videos are available at \href{https://github.com/pairs-lab/WaveSync}{WaveSync}

Diffusion and flow-based generative policies provide a powerful policy class for reinforcement learning by inducing rich stochastic exploration through iterative action generation. However, the stochasticity of diffusion policies is not suitable for stable and precise control in high-dimensional robotic systems, where small action variations can accumulate into inconsistent motion and reduced robustness. To address this issue, we propose SteerGenPO, a latent-space reinforcement learning framework that steers a trained generative policy into a robust deterministic robotic controller. The key idea is to replace stochastic latent sampling of the trained generative policy with a learned latent actor that predicts a state-dependent latent input for the generative policies. This separates exploration and control: stochastic generative sampling provides diverse action proposals during policy learning, while deterministic latent steering provides stable and adaptive control at deployment. We evaluate SteerGenPO on six Isaac Lab benchmarks and a Unitree G1 locomotion task. The results show SteerGenPO improves over both classical RL and generative RL baselines, while its deterministic latent steering produces more stable inference-time behaviors and more reliable command responses.

Registering two captures of the same indoor space taken at different times underpins persistent spatial memory for robots and AR systems, yet the realistic version of this task is egocentric and its most scalable form is RGB-only. Head-mounted cameras yield blurry, fast-moving, partially overlapping views from which dense geometry is hard to recover. Classical registration leans on exactly the clean point clouds this setting lacks, while learned scene-graph methods require a pre-built or annotated graph and a trained matcher that we find brittle under egocentric data. We take a different route, using a pretrained vision-language model as the source of both scene understanding and cross-scan matching. Our method, PROSE (Prompted Scene rEgistration), lifts each RGB sequence into an object-level 3D scene graph using off-the-shelf foundation models for geometry, segmentation, and language, then prompts the same VLM to match object instances across the two RGB sequences. To make this matching tractable and reliable, we leverage object heights as a prior and verify each proposed match with a paired same/different query, then solve for the rigid transform by hypothesizing a candidate per matched object and selecting the one with the strongest geometric consensus. PROSE adds no learned parameters and requires no depth sensor, training, or annotated graph. On the egocentric Aria Digital Twin and Aria Everyday Activities benchmarks, it outperforms both geometric and learned scene-graph baselines in registration accuracy, on ground-truth and RGB-reconstructed point clouds alike, and the scene graph it produces transfers directly to downstream tasks.

Neural ordinary differential equations (neural ODE) have started to appear in safety critical settings such as continuous-time controllers for cyber-physical systems and classifiers integrated into automated decision pipelines, raising the question of whether their behavior can be formally verified. Existing tools dedicated to neural ODE provide only a single reachability call without iterative input set refinement, limiting the precision of their verdicts to whatever one reachability call can deliver. We present TNODEV, the first sound formal verifier for neural ODE that integrates a falsification checker, a fast interval-based reachability backend based on continuous-time mixed monotonicity, a verification and refinement loop with three input-set splitting heuristics, and a parallel scheduler in a single end-to-end pipeline. TNODEV supports safe-set inclusion verification on pure neural ODE, neural ODE in closed loop with a neural network controller and general neural ODE (GNODE), with the safe set specified either as an interval or as the half-space intersection induced by a target classification label. We evaluate TNODEV on a range of benchmarks across safe-set inclusion and classification-robustness properties, including a direct reachability comparison against NNV~2.0 and CORA and a verification comparison against NNV2.0 on MNIST general neural ODE classifiers.

Robotic systems routinely encounter conflicting objectives, modeling errors, and degenerate contact conditions that render quadratic programs (QPs) infeasible. Yet most optimization solvers and differentiable QP layers assume feasibility, leading to numerical failures, unstable gradients, or solver breakdown when constraints cannot be simultaneously satisfied. We present Elastic ODYN, a primal--dual non-interior-point QP solver that handles infeasibility through smooth squared-$\ell_2$ elastic relaxations. The resulting formulation remains well posed under ill-conditioning and degeneracy, supports warm starting, and converges to closest-to-feasible solutions when no feasible point exists. A lightweight refinement stage recovers physically meaningful dual variables from the elastic solution. Building on this framework, we develop Elastic OdynLayer, a differentiable QP layer with stable gradients under infeasibility, and Elastic OdynSQP, an infeasibility-aware SQP method that resolves inconsistent subproblems and intrinsically infeasible optimal control tasks through selective constraint relaxation. We evaluate the framework on benchmark QPs, singular contact mechanics, differentiable parameter identification, and quadrupedal and humanoid trajectory optimization. Across all settings, Elastic ODYN consistently outperforms state-of-the-art elastic QP solvers in robustness, warm-start performance, and convergence reliability, enabling optimization, simulation, control, and learning beyond the feasibility assumptions of existing methods.

Roundabouts challenge automated driving in mixed traffic, as heterogeneous and non-deterministic human behavior, unknown driving intentions, and high interaction complexity create uncertainty about whether the conflict zone will be blocked or available at the moment of entry. We present ROSA-RL -- uncertainty-aware Roundabout Optimized Speed Advisory with Reinforcement Learning. It enables safe and efficient roundabout entry for automated and human-driven vehicles in mixed traffic through probabilistic conflict forecasting. A Transformer-based model predicts conflict zone occupancy over a five-second horizon, capturing multi-agent interactions to anticipate upcoming conflicts and available gaps. The prediction outputs encode uncertainty in future motion and intent, and augment the state of a classical RL framework, enabling uncertainty-aware speed coordination. Evaluated in simulations grounded in real-world data, ROSA-RL can effectively handle uncertainty and outperform a comparable model-based baseline, closing the gap to an ideal setting assuming fully known occupancy while improving traffic efficiency and safety. The source code of this work is available under: github.com/urbanAIthi/ROSA-RL.

Humanoids deployed in human-centered environments must handle force-interactive tasks, where external contacts introduce unexpected disturbances that disrupt locomotion accuracy and stability. Existing learning-based approaches rely on broad domain randomization, task-specific force objectives, or learning-based force estimators from motion history, each of which compromises accuracy, task transferability, or out-of-distribution (OOD) robustness. We present Analytical Disturbance-Aware Policy Training (ADAPT), a framework that equips humanoid policies with a physically grounded disturbance observer. The core of ADAPT is an analytical whole-body disturbance observer that estimates residual force/torque online with the accessible robot dynamics, without requiring force/torque sensors. Fed directly into the policy, the estimated disturbances give the humanoid an explicit, physics-derived sense of external force/torque that can generalize across diverse unseen scenes. Experiments on a Unitree G1 humanoid show that ADAPT achieves accurate disturbance prediction and stronger robustness than a proprioception-only baseline under torso perturbations, standing pushes, and asymmetric hand payloads, with improved velocity tracking even on OOD disturbances. Moreover, ADAPT enables penalizing inferred disturbances at lower-body joints to encourage lighter locomotion.

The Faraday effect is a hallmark of nonreciprocal light-matter interactions and traditionally requires magnetic bias or intrinsically chiral media. Here we introduce a temporal chiral metamaterial in which an effective chiral response is generated entirely by Floquet modulation, without magnetic fields or structurally chiral constituents. The medium is realized by periodically rotating the principal axes of the permittivity and permeability tensors in time. Using a nonlocal temporal effective medium theory derived from Hamiltonian homogenization, we show that the resulting chiral parameter is an odd function of the wavevector, giving rise to intrinsic nonreciprocity despite Onsager-symmetric constitutive relations. This Floquet-induced chirality produces a temporal Faraday effect, in which the polarization plane of a linearly polarized wave rotates continuously in time. The direction and magnitude of the rotation are programmable through the modulation sequence and remain invariant under both spatial and temporal reversal. Our work establishes Floquet-induced chirality as a fundamentally new mechanism for nonreciprocal light control and opens a route to reconfigurable polarization manipulation in time-modulated photonic systems.

Hamilton-Jacobi-Bellman theory implies that the optimal goal-conditioned action depends on the goal only through the gradient of the goal-reaching distance at the current state, yet standard online GCRL still conditions the actor on the raw goal -- a signal that is geometrically uninformative when the goal is far from the data distribution. We propose Direction-Conditioned Policies (DCP), a fully online method that decomposes goal-reaching into two components sharing one InfoNCE representation $ψ$: a subgoal-scoring step that selects a visited state $z_t$ aligned with the final goal $g$ in $ψ_g$, and a direction-conditioned actor that consumes the unit direction $d_t$ and magnitude $r_t$ from $ψ(s_t)$ to $ψ(z_t)$. The two components train jointly, factor cleanly at deployment (subgoal scoring is removed, while direction conditioning remains with $g$ in place of $z_t$), and admit independent modification at the same $(d_t,r_t)$ interface. We prove three results. First, direction sufficiency under HJB: the optimal action under control-affine dynamics depends on the goal only through the value gradient. Second, a quantitative bound showing that, under mild conditions on the learned representation and assuming the scoring rule returns an on-path $z_t$, the actor's conditioning input at training and at deployment coincide up to representation error and geodesic slack. Third, a controllable-subspace characterization of when directional conditioning fails. Across nine environments, DCP improves over Contrastive RL on most final metrics, with the largest gains on manipulation and obstacle-interaction tasks; a qualitative analysis of the learned $ψ$-distance landscape shows the contrastive representation behaves as an online quasimetric encoding environment topology, and the single failure case (AntSoccer) localizes to a learned-gradient pathology that the theory anticipates.

Autonomous fall recovery is a critical capability for quadrotors operating in real-world environments, where collisions or failures may leave the vehicle resting on the ground in an arbitrary attitude. This problem is challenging because recovery must be achieved under limited onboard sensing, in constrained free space, with ground contact, and in the presence of unknown disturbances. In this letter, we present an RL-based framework for autonomous fall recovery of a quadrotor from arbitrary ground attitudes to stable hover using only lightweight onboard sensors. To address severe partial observability and intermittent sensor invalidity, we train a recurrent policy within an asymmetric actor--critic architecture, leveraging an Incremental Nonlinear Dynamic Inversion (INDI) controller to track the policy output. Combined with high-fidelity simulations of motor response and optical flow, the overall training framework significantly reduces the sim-to-real gap. Simulation ablation studies validate the importance of the main design choices, while real-world experiments demonstrate zero-shot transfer and robust recovery under different initial attitudes, wind disturbances, and additional payloads. These results demonstrate that agile quadrotor fall recovery can be achieved without explicit state estimation using only limited and unreliable onboard sensing.

Modern imitation learning methods, including visuomotor and Vision-Language-Action (VLA) policies, typically output high-level action references that are executed by low-level controllers. However, the absence of higher-order reference signals, together with the policy's lack of awareness of the underlying low-level control dynamics during training, inevitably induces an execution gap. As a result, realized actions deviate systematically from policy-commanded ones, with a critical impact on precision-sensitive manipulation. Prior work either modifies the policy architecture or the low-level controller, both requiring intrusive changes to the pretrained policy or packaged controller. This raises a natural question: when the policy and controller are both treated as inaccessible black boxes, can we bridge the execution gap? We propose Adaptive Policy Execution (APEX), a plug-and-play framework inserted between the policy and the controller that reconstructs a dynamically feasible reference from policy outputs and adapts at test-time according to low-level state feedback, with a provable convergence guarantee. Extensive empirical studies show that APEX reduces controller-induced tracking error by 41.2% on demonstration replay and improves manipulation success by 4.8--25.8 percentage points across four visuomotor and VLA policy classes.

Many real-world manipulation scenarios, such as handling complex collaborative tasks and dealing with large workspaces, require coordination of more than two robotic arms. Consequently, an effective multi-arm teleoperation system is required to collect demonstrations for training coordinated multi-arm manipulation policies. However, existing teleoperation frameworks mainly focus on single-operator or multi-operator setups, facing a practical trade-off between the cognitive load placed on a single operator and the coordination cost incurred by multiple operators. To address this problem, we introduce HATS, a human-agent teleoperation system that enables a single human operator, assisted by an MLLM-based agent, to collect data for multi-arm manipulation tasks. Our system decouples the control space: two primary arms are directly teleoperated by the human, while two assistive arms are controlled by a training-free agent that handles sub-tasks. In addition, the human operator can use voice commands to prevent collisions and correct assistive arm behaviors during execution. Extensive evaluations demonstrate that HATS achieves data collection efficiency and success rates comparable to expert dual-human teams. Moreover, downstream policy evaluations demonstrate the efficacy and quality of the data collected through HATS.

Collaborative mobile manipulation requires robots to coordinate with a partially observed partner while physically interacting through shared objects. This is difficult because failures often arise not from poor local skills, but from mistimed waiting, yielding, pulling, releasing, or repositioning. We study this problem with two bimanual mobile manipulators coupled through rigid and deformable objects. We propose Sequential Asymmetric Imitation (SAI), a single-teleoperator curriculum for learning coupled multi-robot behaviors without synchronized dual-operator demonstrations or explicit inter-robot communication. SAI trains Robot A from unilateral demonstrations with a compliant human partner, trains Robot B against the deployed Robot A policy, and then refines Robot A using sparse interventions near coordination failures. This staged process exposes the policies to increasingly realistic partner behaviors, including delay, phase mismatch,insufficient yielding, and interaction conflict. Across real-world dual-robot manipulation tasks, SAI improves task success, phase synchronization, and partner-contingent yielding over independent imitation and curriculum-ablation baselines. These results suggest that physically coupled collaboration can be learned through the structure of the imitation curriculum, rather than through synchronized multi-operator demonstrations or explicit coordination mechanisms.Project page:http://cyc0429.github.io/sai-project-page/

Robots deployed in the real world must plan motions across diverse scenarios without per-scenario retuning. End-to-end reinforcement learning (RL) can generalize across scenarios but often becomes brittle under distribution shift, reward misspecification, and stochastic interactions. Model predictive path integral (MPPI) control enables strong real-time refinement without gradients, but its performance depends on a well-shaped sampling prior, while manually designing the priors does not scale to multi-scenario deployment. We present HOLO-MPPI (High-level Offline, Low-level Online MPPI), a multi-scenario motion planning framework that combines high-level policy learning with low-level stochastic optimal control. Offline, we learn a high-level policy that proposes scenario-robust plans in an abstract action space, with a learned world model for online rollout. Online, the policy serves as a data-driven prior generator that parameterizes MPPI's sampling distribution conditioned on the current observation and goal. MPPI then optimizes low-level control sequences around this prior in real time to adapt to local disturbances. We instantiate HOLO-MPPI in autonomous driving by designing an effective high-level action space and tailored model architectures. Our evaluation across diverse driving scenarios shows that HOLO-MPPI improves upon MPPI and end-to-end RL baselines while maintaining real-time control.

Monocular visual odometry (MVO) is foundational to autonomous navigation and robotic localization. However, existing learning-based MVO approaches often struggle with either a lack of interpretable, complementary features or overly complex multi-stage architectures. These limitations inherently restrict their robustness and cross-domain generalization. In this work, we propose MVOFormer, a novel transformer framework for robust monocular visual odometry. Our architecture features a Flow-Semantic Dual Branch Encoder that synergizes dense geometric motion cues with object-centric semantic priors, explicitly distinguishing static structures from dynamic distractors. These representations are then fused by an Iterative Multimodal Decoder, enabling coarse-to-fine pose refinement while dynamically suppressing attention on unreliable regions. Extensive evaluations demonstrate that, without any target-domain fine-tuning, MVOFormer achieves superior zero-shot generalization and robustness, significantly outperforming prior learning-based frame-to-frame methods across diverse benchmarks including TartanAir, KITTI, TUM-RGBD, and ETH3D-SLAM.

Translating video demonstrations into executable robot commands remains challenging because existing methods often fail to identify which objects are functionally involved in the demonstrated action. As a result, they may generate commands that are linguistically plausible but operationally ambiguous. We propose an object-centric video understanding framework that decouples action recognition from object identification to generate precise, grammar-free manipulation commands. Our approach integrates Temporal Shift Modules (TSM) for efficient spatio-temporal action classification with a novel \textbf{Object Selection} algorithm that identifies task-relevant objects through trajectory-based role classification, blur detection, and overlap minimization. The selected objects are then processed by Vision-Language Models (VLMs) for robust category recognition and zero-shot generalization. Evaluated on a modified Something-Something V2 dataset, our method achieves 86.79\% action classification accuracy and BLEU-4 scores of 0.337 on standard objects and 0.261 on novel objects. These results improve over the strongest task-specific baseline by 80.2\% and 143.9\%, respectively. Larger gains are observed in METEOR and CIDEr, reaching 157.9\% and 171.7\% on novel objects. Across all semantic metrics, our approach consistently outperforms task-specific methods and remains competitive with, or surpasses, large general-purpose VLMs while retaining a modular, object-centric design.

In cyber-physical systems (CPSs), fault tolerance is traditionally achieved by analysing sensor and actuator outputs, detecting progressive drift or sudden failures, and initiating suitable tolerance mechanisms. Reasonable under general failure models, this approach fails to capture nuanced disruptions caused by cyberattacks, which may employ subtle strategies. This is particularly critical in embodied CPSs, where computational and physical devices not only have an active role in task completion, but also in embodiment preservation (that is, maintaining the system's physical integrity). To prevent structural physical damage, embodied CPSs require a framework that enables proactive response to cyberattacks. This paper proposes a formal dependability framework that incorporates IDS information into resilience evaluation predicates, enabling assessment of tolerance to disruption and degradation. The framework supports structured reasoning about how cyberattacks affect task execution and embodiment preservation, and whether mitigation strategies must be deployed. Analytical examples demonstrate its analytical capability and soundness, establishing a theoretical foundation for dependable and secure embodied CPSs.

Code-as-Policies (CaP) has shown that large language models (LLMs) can write code to solve robotics tasks by composing perception, planning, and control primitives. Recent CaP systems, however, rely on multi-turn code-generation loops at test time, which is often infeasible for real-time robot control. We introduce Robotics Harness Optimization (RHO), a novel paradigm in which tool-enabled coding agents, at training time, propose and search for interpretable, neurosymbolic multi-file policy repositories (Repositories-as-Policies) that compose these primitives rather than a single prompt, function, or file. RHO searches with reflective feedback from environment reward and execution rather than teleoperation demonstrations. It generalizes to perturbed pick-and-place settings like LIBERO-PRO, where OpenVLA scores 0.0% and $π_{0.5}$ averages 12.83%. Using the same low-level primitives, RHO reaches a 45.0% success rate, 2.5x higher than the strongest multi-turn agentic system, and 3.5x higher than $π_{0.5}$. On Robosuite, RHO sets a new state-of-the-art of 70.0%, exceeding the prior multi-turn record of 68.29% using single-turn execution with no corrective LLM code edits at deployment. When an LLM is used in the control loop, as on RAI's O3DE benchmark, RHO optimizes the deployed agent's multi-file harness of prompts, tools, and control code, improving held-out success from 23.5% to 44.3% with 20% less wall-clock time and 27% fewer tool calls.

Imitation learning has enabled highly-dexterous robotic manipulation from RGB observations. Policies trained with these methods, however, typically condition robot actions on only a short history of observations. These policies cannot solve tasks that require memory and can get stuck repeatedly executing the same failing motions. In this work, we first benchmark policy performance as context length is incrementally increased from short to long, across a spectrum of tasks with varying local stability and memory requirements, and in multiple data regimes. To our knowledge, this is the first study to investigate context length in imitation learning at this level of detail. Our results challenge prior claims: naively scaling context length is not as brittle as advertised in literature. With an appropriate conditioning method and denoising backbone (UNet+Cross-Attention), single-task policies achieve high success rates on many tasks in the usual data regime even with naive scaling. Next, we propose a training algorithm to jointly train policies at multiple context lengths, further reducing the sample complexity of long-context learning. Finally, we apply our findings to re-evaluate some previously proposed solutions to long-context imitation learning.

Achieving autonomous robotic dexterous manipulation requires precise, human-like action sequences at scale. As a scalable supplement to costly teleoperation data, extracting trajectories with both visual fidelity and physical plausibility from monocular videos represents a promising frontier in embodied AI. To this end, we introduce V2P-Manip, an efficient framework designed to learn dexterous manipulation policies directly from human demonstration videos. We establish an efficient, integrated pipeline encompassing 3D asset acquisition, trajectory estimation, and dexterous policy learning. To bridge the gap between visual perception and physical constraints, we introduce a two-stage refinement process to enforce spatial alignment and physical consistency. Evaluations on the TACO and OakInk benchmarks demonstrate that our approach significantly outperforms previous methods in pose accuracy, adaptability to unstructured environments, and training efficiency. Ultimately, experimental results confirm an average success rate of over 75% across multiple synthetic manipulation tasks and validate the adaptability of the extracted manipulation priors across diverse dexterous hand embodiments.

Optical tweezers enable the manipulation of microscopic objects using light, yet the fundamental limits to the optical forces that can be exerted on matter remain unknown. Here we derive a general upper bound to the maximum optical force that can be applied to a particle, based on an expansion of electromagnetic fields into well-defined helicity multipolar modes. This method finds the optimal force for any kind of fields external to the particle, including evanescent fields. We apply the method to homogeneous spherical particles in a stable trap and identify the field distributions that saturate this bound for the trapping stiffness. We further provide experimentally accessible strategies to approach these optimal conditions, including configurations using counterpropagating and single-beam traps. Experiments demonstrate a threefold enhancement of trapping forces relative to conventional designs, while theoretical predictions indicate that order-of-magnitude improvements are achievable for larger particles and high angular momentum beams. Our results establish fundamental design principles for maximizing optical forces and define the ultimate limits of optical manipulation.

The integration of augmented reality (AR) and EEG-based brain-computer interfaces (BCIs) offers a promising path for enabling intuitive control of robots for assistive purposes. However, existing AR brain-robot interface (BRI) systems are often constrained to task-specific structures, limiting their utility in real-world environments. We present an AR BRI designed for generalist robot arm manipulation that combines gaze-based object selection with motor imagery action control. Our system uses eye-tracking for intuitive object targeting and context-aware visual overlays ("Place" and "Use") to guide the user through tasks within a shared autonomy framework. We evaluated the interface through a feasibility study with 18 healthy participants performing three multi-step activities of daily living: drinking, using a drawer, and operating an oven. Our results demonstrate that this interaction paradigm enables effective sequential task execution and high user engagement, achieving a "Good" usability rating (SUS > 70). These findings support the feasibility of the proposed interaction paradigm for complex BCI-driven robotic assistance, and motivate future evaluation with the intended target population. Project website: https://ar-bri-manip.github.io/.

Learning-based visual navigation has enhanced semantic goal-reaching capabilities. However, due to their black-box nature, purely end-to-end models often lack explicit geometric constraints, leading to unpredictable and unreliable obstacle avoidance in open environments. Conversely, traditional geometric planners ensure safety but struggle with high-dimensional visual targets. To address these limitations, we propose SemGeoNav, a novel hierarchical visual navigation framework.It tightly integrates the high-level semantic reasoning of end-to-end models with the reliable local planning ability of geometry-based methods, achieving robust image-based navigation while significantly improving obstacle avoidance. Furthermore, we introduce a temporal trajectory smoothing mechanism to ensure continuous and stable robot motion. We evaluated SemGeoNav on a Unitree Go2 quadruped robot in real-world environments. The results demonstrate that SemGeoNav outperforms existing representative methods, including ViNT and NoMaD, achieving higher success rates and shorter navigation times.

Low-loss, broadband photonic integrated circuits (PICs) are critical enablers for optical communications, photonic computing, and quantum applications. Lithium tantalate on insulator (LTOI) is an emerging photonic platform offering a wide transparency window and strong Pockels effect, and thereby enabling efficient electro-optic modulation and high data rates. Here, we present the first implementation of efficient out-of-plane polymer coupling interfaces fabricated via 3D direct laser writing for both fully etched strip and partially etched rib LTOI waveguides, achieving ultra-low coupling losses of 0.9 dB (strip) and 1.25 dB (rib) per interface. Both coupler types exhibit a 3 dB optical bandwidth spanning more than an octave from 850 nm to 1740 nm and maintain stable operation under 1 W optical input power. Combined with on-chip waveguides exhibiting propagation losses below 0.1 dB/cm, these characteristics represent a key step toward unlocking the full potential of LTOI for high-speed optical signal processing with an unprecedented degree of parallelism. In addition, the octave-spanning bandwidth enables efficient interfacing of both the fundamental and second-harmonic signals, making the platform highly attractive for second harmonic generation based quantum squeezing applications.

We present ART-Glove, an articulated tactile glove designed to capture contact-grounded dexterous demonstrations while preserving human dexterity. ART-Glove makes hand-side contact geometry explicit with 16 rigid functional surfaces covering the fingers, thumb, and palm. Twenty-two anatomically aligned joints connect these surfaces and allow them to follow human hand motion during dexterous manipulation. Encoder-based sensing tracks surface motion, while dense piezoresistive tactile sensing records contact over the same surfaces. The complete system captures synchronized 22-DoF joint measurements and 2048-taxel tactile measurements at 120 Hz. We evaluate ART-Glove across experiments on motion freedom, joint sensing, tactile sensing, and contact-rich interaction capture, demonstrating its ability to preserve human dexterity while recording contact-grounded information that can support downstream dexterous robot learning.

Controlling mid-infrared (MIR) photons using well-developed telecom photonic platforms would enable new functionalities in molecular and quantum photonics. However, establishing efficient interactions between MIR and telecom photons remains challenging due to their large spectral separation and weak nonlinear coupling. Here, we demonstrate optomechanical control of MIR photons mediated by vibrational polaritons, enabling photon-photon interaction between MIR and telecom fields across distant spectral regions. Using a Fabry-Pérot cavity incorporating a vibrationally active polymer, we observe telecom-driven dissipation enhancement of MIR photons at 9.5 $μ$m with a modulation depth of 1% under a 4 mW pump. The linear power dependence, mixing-ratio dependence, and detuning response consistently indicate a MIR and telecom photon-photon conversion enabled by strong light-matter coupling. This approach establishes a polaritonic optomechanical platform for bridging disparate spectral regimes and provides a dissipation-engineered route toward hybrid MIR photonics and quantum transduction.

Agents increasingly access large language models (LLMs) through API routers. A router terminates the client's transport-layer security session and opens a separate upstream session, so it holds the full interaction in plaintext. This makes the router an application-layer man-in-the-middle: it can rewrite agent tool calls, swap dependencies for typosquatted packages, trigger attacks only under audit-evading conditions, and passively exfiltrate secrets. Existing client-side defenses are evadable. We propose AEGIS, a provider-transparent attested API router whose data path is a client-verified faithful passthrough. AEGISconfines plaintext handling to a small hardware-enclave component while leaving authentication, scheduling, accounting, and management on the untrusted host. The client verifies the enclave before releasing plaintext. The host can neither read nor alter the interaction, and plaintext leaves only toward destinations fixed by the measured image. We show that all four malicious-router attack classes succeed against a plaintext-access baseline and are blocked by AEGIS, including adaptive tests against the same boundary. The trusted path is $851$ lines, carries three provider-native APIs without conversion, and completes every request under real-provider workload and concurrency. In a seeded audit pilot, two commodity coding agents find eight and ten of ten planted invariant violations. The local relay overhead is about six milliseconds per request.

Mid-infrared frequency combs enable highly selective and sensitive molecular spectroscopy by leveraging the strong vibrational transitions in this spectral region. Among these, there is a particular need for compact, tunable sources with electronic control over comb parameters for integrated sensing platforms. In this work, we demonstrate a mid-infrared electro-optic frequency comb source based on nonlinear frequency conversion in thin film lithium niobate. The system combines a near-infrared pump, amplitude-modulated using an integrated Mach-Zehnder modulator for lock-in detection, with a telecom-band electro-optic comb generated via a double-pass phase modulation scheme. Mid-infrared comb generation is achieved through difference frequency generation in a periodically poled waveguide. By tuning the telecom seed laser and the chip temperature, we obtain mid-infrared combs with a bandwidth of approximately 6 nm and center wavelength tunability of over 200 nm. The comb free spectral range is directly controlled via the applied radio-frequency modulation. Operation across multiple integrated photonic circuits reaching wavelengths up to 3.7 $μ$m is demonstrated. Furthermore, dual-tone EO comb generation in the mid-infrared is realized. To our knowledge, this is the first integrated mid-infrared electro-optic comb source offering independent electronic control of both center wavelength and comb spacing.

Space heating in buildings accounts for 10% of the global CO2 footprint. The widespread adoption of energy-efficient heating technology, e.g., heat pumps, could help reduce this figure, but technology alone may not suffice to reach carbon neutrality. Additionally, human occupants have an important role to play by adopting sustainable heating behaviors, e.g., avoid excessive window opening in the winter or (pre-)heat their units while clean energy is abundant. Thus far demand response policies aimed at promoting these behaviors have been monetary, which discriminates against low-income households and exposes human occupants who do not actively engage with real-time control signals to financial risks. This paper instead investigates the suitability of a non-monetary karma economy for promoting sustainable heating behaviors. Karma leverages the repeated and dynamic nature of heating energy allocations to attain climate targets both fairly and efficiently over time without resorting to financial means. As a first step towards experimentally validating the karma concept with real human occupants in the KTH Live-In Lab, we perform a simulation study on a digital model of the Live-In Lab. The study provides initial estimates of expected effects to guide the design of human-in-the-loop experiments, as well as assists with designing and tuning the karma economy in this context. As a specific example, we investigate how incorporating consumption memory in the form of karma affects window opening behaviors in comparison to conventional memory-less heating operation.

This paper addresses real-time state estimation for a nine-compartment nonlinear COVID-19 epidemic model with two co-circulating strains, a super-spreader subpopulation, vaccination with waning immunity, hospitalization, and mortality. Time-varying transmission and vaccination rates are known inputs from a companion calibration, leaving the reconstruction of all nine states from three routinely reported observables: hospitalizations H, fatalities F, and vaccinated stock V. The contributions are theoretical rather than in the filter recursion. First, a Lie-derivative observability analysis yields, via a six-step derivation, the closed-form determinant |det(O9)| = delta_w * gamma_a^2 * kappa * rho2 * w1^2 * (delta_i - delta_p)^2 * |r1 - r2|, showing the level-2 codistribution is rank-deficient at the calibrated symmetric parameters (delta_i = delta_p, r1 = r2); the third Lie derivative restores full rank 9, with r2 the symmetry-breaking parameter. Second, an EKF is designed on the Euler-discretized dynamics with a closed-form 9x9 Jacobian and Joseph covariance update. Third, local exponential mean-square boundedness of the error is proved as a full theorem via the Reif-Gunther-Yaz-Unbehauen hypotheses, exploiting the bilinear drift and linear output to obtain a global-radius quadratic remainder bound that extends to bilinear-drift, linear-output systems. Fourth, the noise covariances are designed from calibration residuals and assessed by NEES and innovation-whiteness tests. All experiments use synthetic measurements from the calibrated model, so reported RMSE values (0.07%-2.72%) are methodology benchmarks, not predictive accuracy. A parameter-mismatch study shows measured and directly-coupled channels stay accurate under model error up to +/-30% while indirectly observed states degrade gracefully. The framework provides the state-feedback basis for future Model Predictive Control.

Flow Matching (FM) is a powerful approach for behavior cloning in multimodal action spaces [Jiang et al., 2025], but because it is not trained to directly maximize expected return, there is still room to improve how FM policies act at test time. This work investigates whether a learned world model can improve FM policies by enabling Model Predictive Path Integral (MPPI) planning over candidate action sequences proposed by the policy. Building on TD-MPC2 [Hansen et al., 2024], I introduce FlowMPC, a framework that combines an imitation-learned FM policy with a learned world model for test-time planning in ManiSkill manipulation tasks [Tao et al., 2025]. Across PickCube and PickSingleYCB, adding the world model improved performance over the FM policy alone, with especially clear gains in end-of-episode success. These results suggest that world-model-based planning can effectively complement flow-based imitation policies without modifying the FM training objective.

This paper studies safe distributed consensus for single-integrator multi-agent systems over connected undirected graphs under simultaneous asymmetric actuator constraints and output safety constraints. Each agent is equipped with a continuously differentiable asymmetric actuator dynamics that maps a commanded control signal to the realized plant input while keeping the latter strictly inside a prescribed admissible interval. To address output safety, a barrier-coordinate transformation is introduced over a common time-varying safe interval, and a distributed synchronization law is designed in the transformed coordinates. The resulting controller integrates a graph-based coordination layer with an actuator-side tracking layer, thereby enabling simultaneous enforcement of input admissibility, forward invariance of the safe output set, and asymptotic synchronization. For compact admissible sets of initial conditions, it is shown that the closed-loop solution is complete, all signals remain bounded, the actuator inputs remain strictly within their asymmetric bounds, and the agent outputs remain inside the prescribed safe interval for all time. Moreover, the transformed synchronization errors converge exponentially to zero, and the original agent outputs asymptotically synchronize to a designer-selected admissible trajectory embedded in the common safe interval. Numerical simulations validate the proposed framework and demonstrate safe consensus under both asymmetric actuation bounds and time-varying output constraints.

Terahertz technology is a key enabler for sixth-generation (6G) wireless networks, yet its application is constrained by increasingly severe free-space loss at high frequencies. To efficiently retrieve weak signals at the receiving end, a compact frontend that features both a high-gain antenna and a low-noise signal-detection chain is critical. Current transistor-based THz electronic frontends face significant challenges in meeting these demands because both on-chip antenna efficiency and transistor noise performance degrade rapidly when approaching their cut-off frequencies. Photonic technology provides an alternative solution to circumvent the transistor bandwidth limit, yet most microwave photonic links to date exhibit noise performance substantially worse than state-of-the-art electronics. Here, we demonstrate low-noise integrated THz photonic frontends that deliver undegraded link noise performance across three major THz windows from 140 to 450 GHz, and outperform electronic frontends in the upper two windows. We achieve this through co-design of high-gain on-chip THz antenna array and broadband THz-optic modulator on a single thin-film lithium niobate (TFLN) chip, leading to distributed reception of free-space THz signals and continuous coherent build-up of the THz-optic conversion process with unprecedented efficiency. Combined with an efficient heterodyne detection chain, our integrated frontends exhibit effective isotropic noise figures of 13.6 and 16.2 dB at 250 and 450 GHz, respectively, both setting new benchmarks in their respective bands. We further demonstrate 6G-oriented multi-link communication up to 20 Git/s. Our integrated frontends represent a significant step towards compact, cost-effective and energy-efficient THz wireless systems in 6G and beyond.

Bit-serial computation enables sequential processing of data at the bit level, providing several advantages, such as scalable computational precision. This approach has gained significant attention, especially for exploiting bit-level sparsity in AI workloads. While current bit-serial processors leverage bit-level sparsity to eliminate the computation associated with zero bits, they face a fundamental trade-off: either they suffer from low memory-access and computation efficiency caused by irregular patterns of non-zero bits, or they incur substantial area overhead from complex online scheduling mechanisms required to reorganize bit-level data and preserve memory access and computation regularity. Therefore, we present the SparseCol processor, designed to harness extensive bit sparsity while maintaining high hardware utilization across various AI applications, including CNNs, RNNs, and transformers. In contrast to traditional methods, SparseCol exploits structured bit-level sparsity, denoted by bit-column sparsity, without requiring any re-training. Furthermore, SparseCol implements a dynamic dataflow architecture that tackles hardware under-utilization issues commonly found in existing bit-serial solutions. Fabricated in 16nm CMOS node, SparseCol delivers 1320 BTOPS/W (BTOPS represents Binary Tera-Operations Per Second, calculated as #W bits x #A bits TOPS) peak efficiency while maintaining accuracy, outperforming SotA sparse processors in terms of efficiency by 6.8x. Comprehensive evaluations on CNN classification tasks and transformer architectures demonstrate system-level efficiencies of 745.02 BTOPS/W and 850.5 BTOPS/W, respectively.

Many games rely on storytelling combined with systems that track levelling, NPC behaviour, and consequence simulation; bridging tightly-authored narrative with deeply-simulated worlds -- most acute in sandbox and open-world settings -- has been prohibitively expensive. LLM-driven worlds open a new path: a single harness can coordinate numerical state, narrative voice, storytelling pacing, and rule logic together. Realising this requires the LLM system to sustain a persistent world (who is where, what has just happened, what is currently true), which today's deployed systems do not: the narrative voice asserts state in free prose without any validated representation, so a fully autonomous game engine remains infeasible. We treat this as an architectural choice, not a limitation of language models, and report work in progress on a framework -- orchestrated reality -- that makes the world a canonical object owned by a singleton orchestration agent analogous to the tabletop-RPG Game Master (GM). We formalise an LLM-driven game world for a human player as a Parameterized-Action POMDP: state is a tree of canonical JSON entities, actions decompose as $a=(k, x_k)$ (a discrete intent kind plus structured JSON parameters), the agent observes only a narrative projection $o=O(s)$ of state, and the transition kernel $F$ is an LLM-driven Plan-Diff-Validate-Apply (PDVA) pipeline that commits schema-validated, content-hashed JSON deltas. We give the formal model, a JSON-state example, a worked single-turn example, and a catalogue of 15 illustrative incidents drawn from a real deployment showing the framework in action. Empirical validation through a planned human player study -- together with multi-NPC concurrent agency and deployment as an RL environment -- is situated as future work.

Power-converter-rich electrical systems, formed by renewable generation, electrified transportation, and inverter-based resources, exhibit strongly nonlinear dynamics, multi-physics design tradeoffs, fast control requirements, and growing reliability and cybersecurity constraints. These characteristics strain workflows that rely only on physics-based modeling, sequential optimization, and rule-based operation. This paper reviews artificial intelligence (AI) for power-converter-rich electrical systems through a life-cycle and deployment-readiness perspective. The literature is organized across converter design, real-time control, system-level operation, and compliance-oriented governance. For design, we examine surrogate modeling, topology and parameter synthesis, EMI/EMC-aware optimization, reliability-oriented design, and knowledge-assisted workflows. For control, we compare supervised learning, reinforcement learning, learning-augmented predictive control, and safety-constrained learning according to their role in closed-loop implementation. For operations, we focus on microgrid coordination, forecasting, distribution-system observability, privacy-preserving coordination, and cyber-resilient operation where converter-interfaced resources shape the operating problem. Across these stages, the review emphasizes deployment-critical gaps, including stability certification, constraint satisfaction, interpretability, extrapolation, data efficiency, sim-to-real transfer, embedded latency, cybersecurity, privacy, and standards alignment. The resulting taxonomy is intended to clarify where AI is already useful as an engineering support tool and where further validation is needed before autonomous or safety-critical deployment.

Historical medical archives and traditional medicines hold immense potential for drug discovery and remain a primary source for current drug development. However, pre-ontological prose and idiosyncratic taxonomies prevent the standardization and medical modernization of the data for use in current biomedical pipelines. Furthermore, no existing LLM agent system, whether tool-calling, retrieval-augmented, or agentic deep-research, can convert such text into verifiable drug-discovery leads at scale. We close this gap with DeepRoot, a multi-agent LLM system that jointly builds and utilizes a verified knowledge graph, showing that grounding and reasoning -- often conflated -- are separable axes the system can compose for therapeutic reasoning. Applied to the Shen Nong Ben Cao Jing, DeepRoot recovers $10$ of $21$ held-out compound-disease treatment pairs at R@$20$ ($47.6\%$ vs $4.8\%$ for a raw corpus LLM and $\sim\!2.4\%$ random) and dominates an LLM-as-judge audit for reasoning quality over baseline LLMs and LLMs with direct tool-call access to the same APIs DeepRoot itself queries. Tool-using LLMs hallucinate evidence on $87\%$ of claims, versus 7-10% for DeepRoot. Graph-only inference hallucinates $0\%$ but ranks lowest on reasoning coherence; DeepRoot KG+LLM is the only condition to win on both axes, pointing toward a route for systematic mining and repurposing of historical medical knowledge.

Open-source LLM agent ecosystems are growing rapidly, yet the security of community-contributed skills - modular tool definitions that extend agent capabilities - remains largely unvetted. The gap we fill: existing scanners operate at the code layer and are structurally blind to instruction-layer and multi-agent risk - natural-language directives that hijack an agent, exfiltrate data through encoded side channels, or chain harm across pipelines - so what is needed is a semantic, multi-dimensional vetting system rather than another signature matcher. We present SKILLVETBENCH, a live public leaderboard on Hugging Face that uses an LLM-as-Judge to vet agent skills. What is new: SARS (Skill Agentic Risk Score), a five-dimensional agentic-risk metric with a principled weighted formula for instruction-following systems. What is integrated: full CVSS v4.0 vector decomposition and a ClawHub dual-view that places our LLM-generated review beside the official marketplace verdict. What is demonstrated: drawing on our companion benchmark paper [ 1], the LLM-as-Judge stage achieves zero false negatives across 78 confirmed-malicious skills and zero false positives across 22 benign controls, while the best static baseline (SKILLSIEVE) still misses 15%; for instruction-layer categories such as Prompt Injection and Memory Poisoning, conventional tools miss between 89% and 100% of threats (e.g., CODEBERT detects none of nine memory-poisoning skills). Detection rates vary from 35% to 95% across four LLM evaluators, motivating ensemble scoring in production deployments.

The computer systems community has recently seen growing interest in AI-driven system evolution, where AI agents iteratively rewrite systems. Frameworks such as AdaEvolve and Engram report 12-60% score improvements over human-designed algorithms. While these results are promising, there are practical concerns if these AI-evolved programs can perform worse on unseen workloads and exhibit scalability regressions. Given the speed and scale of AI-generated code, we need automated mechanisms to uncover such identify hidden weaknesses in AI-evolved systems programs. To this end, we develop AIChilles that takes as input a baseline program $P$ and an AI-evolved program $P'$, AIChilles searches for valid workloads where $P'$ regresses relative to $P$ in correctness, runtime, memory usage, or output quality. To tackle the diversity in system applications, weakness types and potential bugs, AIChilles combines deterministic workload-parameter extraction, agent-based constraint inference, differential oracles, and code-frequency coverage to discover diverse failures. Across five system applications and 30 AI-evolved programs, AIChilles finds 49 distinct hidden weaknesses. We also show that explicitly including AIChilles in the AI-driven development lifecycle can mitigate several of these weaknesses.

Optical backflow in paraxial Gaussian beams is investigated within the Maxwell framework. Scalar potential representations are employed to identify conditions under which the longitudinal Poynting component becomes negative, showing that backflow is enabled by local suppression of the leading-order transverse field and the dominance of higher-order vectorial contributions. The spatial topology of backflow regions is shown to be governed by polarization through the number of independent local constraints on the transverse field. When the local polarization phase is free, as in the generic case of circular polarization, the leading-order field vanishes only at isolated points, giving rise to point-like backflow regions (extended curves may arise if an additional global phase constraint is imposed). In contrast, when the polarization phase is locally fixed, as for linear, radial, or azimuthal polarization, the suppression condition reduces to a single real constraint, resulting in extended backflow curves. Analytical Gaussian-polynomial solutions explicitly illustrate these effects. These results clarify the role of vectorial interference, establish a polarization-controlled backflow geometry, and provide a foundation for further studies of optical backflow in structured and nonparaxial beam configurations, as well as potential applications in optical manipulation and structured light design.

A theoretical and numerical study of optical forces acting on a Rayleigh particle in a paraxial Gaussian light beam exhibiting regions of optical backflow is presented. Within the dipole approximation, the total optical force is decomposed into gradient, scattering, and spin-curl terms. Vector fields satisfying the exact paraxial Maxwell equations are employed to describe the structured light configuration responsible for two distinct mechanisms leading to backward optical forces. The first originates from the local reversal of the Poynting vector, which induces a negative longitudinal momentum flux, while the second arises from the spin-dependent component of the force associated with the spatial variation of the optical spin density. Analytical expressions and numerical simulations confirm that both mechanisms can produce backward motion of a Rayleigh particle under appropriate beam conditions. These results provide a unified physical picture of backward-directed optical forces in Gaussian beams and open possibilities for particle manipulation in structured light fields.

Traffic scene understanding requires models to reason beyond object recognition, including lane topology, multi-view geometry, temporal evolution, and signal-phase semantics. However, existing traffic-oriented multimodal benchmarks largely emphasize passive visual recognition or isolated video understanding, offering limited support for evaluating structure-aware traffic reasoning under controlled conditions. We introduce OmniTraffic, a controllable generation pipeline and benchmark for spatio-temporal traffic reasoning. Built around 12 real-world intersections reconstructed into editable 3D traffic environments and complemented by surveillance footage from two countries, OmniTraffic supports both controlled and natural-condition evaluation. It defines a three-level task hierarchy spanning scene perception, multi-view and temporal reasoning, and decision support. Using structured traffic metadata, OmniTraffic generates synchronized multi-view VQA samples covering vehicle states, lane functions, view--BEV correspondence, temporal dynamics, and signal-phase analysis, resulting in 8M VQA samples and a 3K human-verified test set. Evaluation of eleven frontier MLLMs reveals a large human--model gap, with the most pronounced failures in topology-grounded and spatio-temporal reasoning tasks. Fine-tuning a lightweight MLLM on simulated OmniTraffic data further improves performance on real-world traffic scenes, demonstrating the value of simulation-generated supervision for traffic-specific multimodal reasoning. Beyond a fixed dataset, OmniTraffic provides an extensible pipeline with configurable intersections, camera views, traffic demands, signal phases, visual conditions, and rare events.

Jordan faces severe water scarcity with 50\% of water produced is lost to leakage, theft and metering issues also known as non-revenue water (NRW). Traditional reactive approaches have proven insufficient for sustained NRW reduction. This paper proposes an intelligent framework integrating EPANET hydraulic modeling, digital twin technology, SCADA systems, and large language model (LLM)-based AI agents for continuous network monitoring and adaptive decision-making. The system combines real-time data streams with physics-based simulation to detect anomalies, employing retrieval-augmented generation (RAG) for policy interpretation and function calling for network control. A proof-of-concept implementation validates technical feasibility using EPYT with offline LLMs (llama3.1:8b via Ollama) on a 1,164-junction Amman district network. The system demonstrates automated hydraulic simulation, flow-based anomaly detection aligned with water distribution zone (DZ) practice, and AI-generated health reports with response times under 2 minutes and zero API costs. Burst detection relies on local flow anomaly analysis: a 30.1~L/s simulated leak produces measurable flow redistribution in 15 pipes, flagging a 15-junction cluster that localises the burst -- confirming alignment with water distribution zone (DZ) monitoring practice. The framework accommodates Jordan's intermittent supply patterns and limited automation through phased implementation, offering a scalable pathway for water-scarce regions to leverage intelligent automation for NRW reduction and operational efficiency.

Optical skyrmions, which are the topologically protected quasiparticles and characterized by the nontrivial polarization textures, have emerged as a promising candidate due to their potential applications in optical communication, data storage, and particle manipulation. In this article, we propose and experimentally demonstrate an efficient and tunable approach for the dynamic transformation of generalized optical skyrmionic textures through the interaction of structured vector vortex beams with chiral media. By controlling the enantiomeric excess of an optically active material, we achieve on-demand conversion among Bloch, Neel or any intermediate skyrmionic states, extending also to optical bimerons. The topological conservation of the skyrmion number proves its robustness towards even higher-order textures. While maintaining a common path and stable setup, the proposed methodology provides an efficient and cost-effective approach towards the flexible manipulation of the topological textures, paving the way towards the understanding of topological transformation and engineering optical skyrmions for information processing or particle manipulation.

We present SLS^2, a framework for safe feedback motion planning from pixels using robust model predictive control (MPC) in learned latent world models. Our approach trains an action-conditioned joint-embedding world model with compact Markovian latent states, enabling efficient gradient-based trajectory optimization through learned latent dynamics. To enforce safety for the true system despite imperfect latent predictions, we inform a GPU-accelerated system level synthesis (SLS) robust MPC scheme with conformal prediction to obtain calibrated latent error bounds and robust latent-space constraint sets. We further learn and conformalize a latent constraint checker, allowing the SLS planner to impose probabilistic safety constraints during closed-loop execution. We evaluate our method on vision-based control tasks, where it improves both goal-reaching performance and safety over latent world-model and safe-planning baselines.

Generating high-quality Physics Word Problems (PWPs) that are novel, complex, and solvable remains a challenging and underexplored problem in educational content generation. Existing approaches, many adapted from Math Word Problem (MWP) generation, often produce ambiguous, unsolvable, or structurally simple questions with limited linguistic diversity. We introduce ARVRE (Agentic Retrieval Value Reinforced Equation-chain), a two-stage framework for generating diverse and mathematically valid PWPs. In the first stage, a form of offline temporal-difference learning is used to construct valid chains of physics equations, while an agentic retrieval-augmented generation (RAG) framework dynamically selects topic-specific concepts and vocabulary. This design enables explicit control over problem structure and difficulty. In the second stage, a Large Language Model (LLM) converts the equation chain and retrieved concepts into a natural-language physics question. By grounding generation in valid equation chains, our method preserves mathematical correctness while promoting linguistic diversity and contextual richness. Human and automated evaluations demonstrate that ARVRE generates PWPs that are more complex, novel, and solvable than those produced by existing approaches. These results highlight the potential of combining reinforcement learning, retrieval, and LLMs for reliable generation of educational physics content.

In this work, we present a simple, exact, and fully analytical method for generating light sheets parallel to the propagation direction, with amplitude and phase envelopes structured on demand. We validate the approach theoretically and experimentally by imprinting images onto light sheets, and we compare the theoretical performance with that obtained using an alternative strategy based on arrays of Frozen Waves (FWs). In this context, the proposed method provides a more direct and flexible control of the field envelopes on the light sheets, resulting in higher-fidelity reconstructions than those achieved with FW-based approaches. The method thus offers a versatile framework for structured light-sheet generation, with potential applications in optical manipulation, microscopy, and 3D holographic imaging.

Discovering the governing equations of a physical system from data is a central goal across the sciences, yet in most experiments only a few states are accessible while the rest stay hidden. Existing approaches treat this partial observability as an obstacle to be removed by first reconstructing the hidden state -- a step that is ill-posed under noise and that discards the physical constraints, such as energy conservation, that the true dynamics obey. We show that for conservative (Hamiltonian) systems no reconstruction is needed: projecting the dynamics onto the measured coordinates yields a memory kernel that we prove to be a lossless positive-real rational matrix, whose poles are the hidden natural frequencies and whose positive-semidefinite residues encode the couplings. The governing equation -- and the underlying Hamiltonian -- can therefore be read directly from the autocorrelation of the measured signal, with guarantees of uniqueness and physical passivity, and without neural networks. We validate the approach on linear, nonlinear, and chaotic systems under realistic noise. By recovering interpretable equations of motion that conserve energy by construction from partial measurements, the method offers a common tool for problems spanning mechanics, fluid and plasma physics, and beyond.

Large language models (LLMs) have facilitated impressive progress in software engineering, code generation, tooling, and systems. Concurrently, a significant body of research has developed which explores a growing variety of methods and systems for applying LLMs to hardware and chip design (e.g., systems for RTL code generation based on functional description). However, when it comes to open Verilog/RTL code-generation, we need high-quality training samples to build specialized and more effective LLM systems through fine-tuning or low-rank adaptation. Here, we propose a ``judge-renew-check-renew-check'' (JRCRC) pipeline which updates a current public dataset using a hierarchy of state-of-the-art commercial LLM models differing in their costs and capabilities in RTL code generation. This approach achieves a cost-effective mechanism for filtering and refining code-generation samples into a higher-quality training dataset. Our experiments also identify some common weaknesses of LLMs in rule-based reasoning and logic, and consequently, in RTL code-generation. Having identified these weaknesses, we develop an architecture for incorporating pre-processing tools to dynamically assist the LLMs in inferring logical relationships from tabular data formats. With our tools-assisted architecture for RTL code generation, we achieve significant overall performance gains in the VerilogEval benchmark and outperform many state-of-the-art methods. Our LLM4RTL system achieves performance comparable to that of GPT-4O using a significantly much smaller LLM.

Dexterous teleoperation requires precise arm-hand coordination, low-latency feedback, and robust interaction in real-world contact-rich environments. This paper presents a modular bilateral teleoperation framework that integrates operator-side input interfaces with a robot-side dexterous hand and compliant robotic arm in a unified control architecture. The system supports position-based hand retargeting, differential arm control, multi-scale haptic feedback, and shared control for stable manipulation. We validate the framework through a real-world dexterous manipulation task, highlighting coordinated arm-hand control and contact-aware interaction. Beyond feasibility, we identify key design insights related to cross-embodiment mismatch, haptic feedback granularity, and shared control. The proposed platform provides a practical teleoperation system and a foundation for collecting high-quality demonstrations for future learning-from-demonstration research.

Liquid cooled supercomputers dissipate tens of megawatts of waste heat through cooling plants organized as parallel subloops that serve coolant distribution units. The number of subloops and the assignment of units to them are design decisions fixed at construction, yet they have not been systematically optimized for facilities at this scale. As electricity grids decarbonize, embodied carbon becomes a larger share of facility life cycle emissions and the cost of an unnecessary subloop becomes harder to justify. We present a framework that integrates operational energy from a validated control optimizer based on sequential least squares programming, embodied carbon from a bill of materials, and expected unplanned downtime from a per subloop reliability model. The framework is applied to the Frontier supercomputer, evaluating all 611 ways of partitioning its 25 coolant distribution units into two through six subloops. The life cycle cost and carbon optimum is found at two subloops holding 14 and 11 units, achieving 3,320.7 tonnes of carbon dioxide equivalent and $3.99 million over a seven year horizon, a saving of 50.2 tonnes and $100,000 compared to built four subloop configuration. The optimum remains on the Pareto front in all 15 scenarios of a one at a time sensitivity sweep. A semi-analytical decision rule generalizes the result, predicting four subloops for Aurora, two for El Capitan, and one for LUMI. When reliability is treated as a hard constraint set by operations policy, the four subloop Frontier deployment is consistent with the constrained optimum.

To address the urgent need for multi-channel high-speed electrical interfacing of thin-film lithium niobate (TFLN) photonic circuits, we realize a flip-chip packaging platform capable of simultaneously delivering 13 high-speed and 32 low-speed electronic signals to a centimeter-sized TFLN chip. The platform exhibits low flip-chip bonding loss and low inter-channel crosstalk over a broad bandwidth up to 50 GHz. Leveraging this packaging platform, we demonstrate high-speed electrical interfacing with two proof-of-concept TFLN photonic circuits, namely a 2x8 optical switch and an electro-optic comb-based transmitter. The switch achieves arbitrary 8-channel routing with ~3 dB insertion loss, < -20 dB crosstalk, and an equipment-limited switching time of <= 34 ps. The transmitter circuit includes a 50 GHz electro-optic comb generator with 2.8-dB flatness, a tunable microring to arbitrarily filter one comb line, and a modulator for data transmission at 20 Gbit/s. The packaging platform could significantly advance large-scale TFLN circuits in optical communications, microwave photonics, and photonic computing.

Multi-agent LLM systems -- coding agents, devops agents, document agents -- now routinely run several agents in parallel against the same git tree, Kubernetes cluster, or document. As soon as two of them mutate shared state, they enter the regime classical concurrency control has studied for decades, but classical mechanisms fit LLM agents poorly. A single agent transaction spans minutes of inference, read sets are broad and opaque rather than statically inferable, and the live state agents act on admits neither fork nor buffer, so writes take effect the moment they execute. Locks block long inference intervals; OCC abort-and-retry discards minutes of work on every conflict. This paper builds concurrency control on a capability classical transactions lack: the LLM inside each agent can judge whether a conflicting write invalidates its plan, and can repair exactly the operations that depended on it. Control therefore turns advisory: the runtime informs, the agent repairs. Our protocol, MTPO (Monotonic Trajectory Pre-Order), fixes a serialization order at launch, serves each read the order-filtered value, and applies writes speculatively in place; a one-way notification asks an affected reader to re-judge and patch its plan, while the framework mechanically undoes and reorders misplaced writes through the saga-style inverse each tool registers in advance. At quiescence the run is serializable in the pre-decided order. We realize MTPO as CoAgent, toolcall middleware whose privileged ToolSmith grows footprint-declared, undoable tools online. On ten contended workloads, CoAgent stays within 5\% of serial correctness at a $1.4\times$ speedup and near-serial token cost, where 2PL and OCC surrender nearly all concurrency gains; on a bash-only target system, it grows a 25-tool library online and lifts the task pass rate from 45/71 to 63/71 at $0.80\times$ the time and $0.86\times$ the cost.

Research on optical rotational manipulation leveraging the imaginary Poynting momentum (IPM) force has predominantly centered on cylindrically polarized Gaussian and annular beams. Here, we extend this framework to tightly focused cylindrically polarized structured light fields possessing closed or open arbitrary-intensity trajectories. We systematically elucidate the rotational dynamics induced by IPM in such fields, characterizing the underlying mechanical effects and trapping flexibilities. Notably, despite carrying zero net angular momentum, these fields drive microparticles into bidirectional rotation along predefined trajectories, thereby challenging conventional paradigms reliant on spin or orbital angular momentum. This IPM-mediated optical spanner offers unprecedented spatial degrees of freedom, paving the way for high-precision optical manipulation along arbitrarily tailored paths.

Optothermal tweezers enable the manipulation of a wide range of nano-objects through optically induced depletion forces. Despite significant advances, the temporal dynamics of optothermal trapping remain elusive, as existing methodologies rely almost exclusively on time and ensemble averaging. Consequently, stable trapping cannot be distinguished from local transient accumulation, where the time-averaged concentration increases but particles exhibit rapid, dynamic motion in and out of the trap. Here we investigate optothermal trapping with single-nanoparticle-level analysis and sub-millisecond temporal resolution. Our data resolve the elusive dynamics of 40 nm polystyrene nanoparticles trapped within depletion force potentials in polyethylene glycol solutions, enabling to differentiate the conditions leading to extended trapping times from those leading to transient localization. Numerical simulations corroborate our experimental findings, elucidating how the interplay between thermophoresis and diffusiophoresis governs nanoparticle dynamics. These insights deepen our mechanistic understanding of optothermal trapping and unlock opportunities for single-molecule studies, nanoscale assembly, and targeted drug delivery.

This paper is concerned with the differentially private consensus problem for discrete-time multi-agent systems with communication delays. The purpose of the paper is to achieve differentially private consensus for such systems while protecting the entire delayed initial histories of all agents. A novel adjacency relation for delayed histories is introduced, and a Laplace-noise-based privacy mechanism is developed, where the noise variance is allowed to vary with time and even increase. By using the difference resolvent function method, decay estimates for the fundamental solutions of the delayed difference equations are derived. Based on these estimates and a backstepping technique, mean square weak consensus, mean square strong consensus, and almost sure strong consensus are established. The estimates for the fundamental solutions are also used to derive an explicit sensitivity bound. Furthermore, a constructive parameter design is provided to achieve a prescribed infinite-horizon $ε^\star$-differential privacy level. Numerical simulations illustrate the theoretical results.

Quantum computing has rapidly emerged as a powerful paradigm for tackling computationally demanding problems. In particular, quantum optimization shows strong promise for hard combinatorial problems in power systems, where increasing distributed energy penetration heightens the need for intentional islanding to maintain grid reliability and resilience. However, power system islanding is an NP-hard combinatorial optimization problem that becomes computationally prohibitive for classical solvers as network size grows, motivating the use of quantum computing as a promising alternative pipeline. This study develops a resource-efficient hybrid QAOA islanding framework that brings physics-constrained power-system partitioning into the quantum optimization workflow. The framework combines coherency-informed graph reduction, physics-aware constraint modeling, and structured post-processing to efficiently convert shallow-circuit QAOA samples into high-quality feasible islanding decisions without deep circuits or large shot budgets. The proposed framework is validated on the standard IEEE benchmark systems (9-, 14-, 24-, 30-, 39-, and 57-bus), demonstrating that the hybrid workflow achieves Gurobi-optimal solution quality with a clear quantum resource advantage over vanilla QAOA, while the resulting islanding solutions satisfy all physical feasibility requirements after network separation. This study establishes QAOA-based islanding as a viable quantum approach for critical infrastructure, with structured post-processing as the key enabler of quantum resource efficiency.

Large language model (LLM) agents are increasingly deployed in multi-agent systems where they must coordinate and agree on shared decisions. We ask whether classical resilient consensus theory, developed for deterministic agents, transfers to LLM agents that may behave adversarially. Framing LLM agreement as a Byzantine consensus game, we run controlled experiments on complete and general communication graphs. We find that prompted LLM agents fail to reach agreement that is achievable in principle: consensus can fail even in settings where classical theory guarantees that a convergent algorithm exists, and this failure persists across temperatures and horizons. At the same time, wrapping the agents with classical resilient consensus filters improves agreement. The benefit of filtering depends on how much robustness the underlying topology already provides. Our results suggest that classical resilient consensus theory is a useful lens for the safety of agentic AI.

Complex cognitive, emotional, and social processes shape human evacuations during natural disasters. Accurate modeling and understanding of human behavior in disasters or emergencies can greatly impact the evacuation process by informing more effective planning and resource allocation. However, collecting human data in these situations is very difficult, and existing computational evacuation models assume rational, homogeneous behavior, leading to unrealistic, overly optimistic predictions. To address this gap, we present a simulation framework of sequential human decision-making during an evacuation scenario, introducing cognitively grounded, persona-driven agents. Our framework models evacuation behavior in a grid-based urban environment that evolves over time, capturing fire and other hazards. Human agents are modeled as personas that make sequential decisions in response to environmental stimuli with cognition structured in three levels: high-level evacuation goals, mid-level route reasoning, and low-level navigation. Decision-making is driven by large language models (LLMs) coupled with a cognitive module and calibrated with empirical human evacuation data. We propose a dynamic, stimulus-driven disaster simulation framework that models human evacuation decision-making using persona-conditioned LLM agents and a cognitive hierarchy.

Controlling batch polymerization is challenging because the absence of a steady operating point prevents standard linearization; the dynamics are intrinsically nonlinear; and multi-phase operation induces state-triggered switching. This study systematically combines four established real-time NMPC ingredients, smooth mode blending, advanced-step warm starts, variable scaling, and a capped iteration budget, to attain real-time feasibility without ad hoc switching heuristics. We provide practice-oriented guidance for selecting smoothing gains and locating switching surfaces, and we make explicit the approximations introduced by smoothing such that, with appropriate tuning, the smoothed and original switching logic are numerically indistinguishable at solver-tolerance levels. All results are obtained in closed-loop simulation using an industrial gas-liquid polymerization benchmark with estimator-in-the-loop, compared against PID and conventional NMPC baselines. Results show improved constraint satisfaction and shorter batch duration under bounded computation, while an ablation study quantifies the specific contributions of each component individually

We study controller and control-architecture co-design for dynamic output-feedback systems. The architecture selects active sensors and actuators, sensor-to-actuator links, and link delays, with costs for hardware activation and communication latency. Direct optimization over controller transfer matrices and discrete links is mixed-integer nonconvex; common alternatives fix the architecture, use regularization, or restrict the controller information pattern to a quadratically invariant (QI) class. We instead optimize finite-horizon output-feedback system-level synthesis (OF-SLS) responses. Binary variables select sensors, actuators, links, and delays, and indicator constraints zero unavailable FIR response blocks before the selected delays. For implementation-local OF-SLS architectures, this gives an exact mixed-integer convex program over a prescribed finite delay menu. A global solve certifies the best architecture-response pair for the chosen delay menu, FIR horizon, admissible architecture set, and scalarization weight. The same encoding gives a QI controller-support reference problem. In a vehicle-platoon benchmark, 99 of 8748 architectures are QI-compatible. At equal architecture cost, the selected non-QI OF-SLS architecture reduces performance loss by a factor of 3.8 relative to the best QI architecture and outperforms regularization-based and canonical information-flow baselines.