Physical AI Brief

Cross-modal alignment (CA) and cross-modal prediction (CP) are the dominant paradigms for multimodal representation learning, yet there is no systematic understanding of when each succeeds, when each fails, and when cross-modal training helps at all -- a gap that leaves practitioners, especially in scientific domains like biomedicine or astrophysics, with heterogeneous instruments and multiple levels of organization and measurement, unable to diagnose why standard methods underperform the best single modality. We develop a unified linear framework that addresses both questions. Under a spiked signal-plus-noise model with structured cross-modal nuisance correlation, we derive separation ratios for both objectives that expose complementary failure modes: alignment whitens each modality and fails when nuisance is strongly correlated across views; prediction encodes whatever is cross-predictable through a one-sided whitening, with recovery governed by source-modality quality. The resulting phase diagram partitions multimodal problems into four regimes: Both, CA only, CP only, and Neither. We present a data-driven procedure to locate real-world datasets in this diagram using a small labeled subsample, identifying the preferred objective and prediction direction before any cross-modal training. Experiments on synthetic data, stereo-vision benchmarks, image-caption pairs, and real astrophysical data validate the predictions in the nonlinear regime, including the Neither regime where cross-modal training is actively harmful. Our framework lets practitioners diagnose their multimodal problem and choose the right objective before committing to training. Code to reproduce the results is available at https://github.com/IlayMalinyak/mm_align_vs_pred.

Autoregressive video generation has emerged as a powerful paradigm for World Action Models (WAMs). However, existing approaches suffer from slow training convergence and limited converged accuracy, particularly at high frame rates, as the training supervision is confined to the current chunk without explicit signals about future dynamics; they also suffer from slow inference due to iterative video denoising. In this paper, we present Next Forcing, a multi-chunk prediction (MCP) framework for causal world modeling that enables faster training, higher accuracy, and accelerated inference. Inspired by multi-token prediction in large language models, Next Forcing introduces an MCP training objective that augments the main model with lightweight auxiliary MCP modules to simultaneously denoise video chunks at multiple future temporal horizons (next$^1$, next$^2$, next$^3$ chunks). These MCP modules form a causal chain across prediction depths, where intermediate features fused from multiple layers of the main model are leveraged to predict future dynamics, allowing near-future predictions to inform farther-future ones and providing dense multi-scale temporal supervision back to the main model. During training, the MCP modules significantly accelerate convergence and improve converged accuracy, especially at high frame rates: at 50 fps, Next Forcing achieves a 93.1% relative improvement over LingBot-VA at 5k training steps and 2.3x faster convergence, and establishes new state-of-the-art results on the RoboTwin benchmark (94.1/93.5% on Clean/Random). At inference, the MCP modules can be retained to predict the next video chunk in parallel with the current one, achieving 2x inference acceleration. Next Forcing also demonstrates significant improvements on PhyWorld, a benchmark evaluating adherence to physical laws in video generation, and over 50% FVD reduction on general video pretraining.

Contact-rich manipulation requires robots to continuously perceive and regulate evolving physical interactions under dynamic contact transitions or complex surface geometries. Recent imitation learning methods improve contact-aware control by incorporating tactile or force feedback, but they rarely model the asymmetric spatiotemporal roles of global force and local tactile sensing. To address this, we propose TacForeSight, a lightweight force-conditioned tactile foresight framework for real-time manipulation. The core component is TacForceWM, a tactile world model that predicts short-horizon tactile latent dynamics from dual-finger tactile observations conditioned on high-frequency wrist force and torque signals. Another key component, the Predictive Tactile-Conditioned Policy, leverages the predicted latents as anticipatory contact priors, models the current-to-future tactile evolution via cross-attention, and adaptively fuses visuo-tactile features through a tactile-guided gating module. By forecasting purely within a compact latent space, TacForeSight enables proactive contact reasoning with efficient real-time inference suitable for high-frequency manipulation control. Real-robot experiments on five representative tasks and three in-process perturbation settings show that TacForeSight consistently outperforms existing baselines, particularly under dynamic contact disturbances. All models and datasets will be made publicly available on the project website at https://tacforesight.github.io/ProjectPage.

In this paper, we propose EEVEE, the first multi-dataset test-time prompt learning framework for LLM agents, enabling test-time prompt learning under real-world task streams. Existing methods are largely designed for single-dataset settings, while real-world applications require models to handle heterogeneous input streams drawn from multiple datasets, domains, and task distributions, limiting their practical applicability. To mitigate cross-dataset interference, EEVEE introduces a router that partitions incoming inputs into task clusters and assigns them to suitable prompt configurations. This design is optimized via a router-prompt co-evolution strategy, which employs interleaved router and prompt learning phases to address their mutual dependency. Experiments across multiple datasets demonstrate that the framework improves robustness under heterogeneous data streams while maintaining single-benchmark learning capability and efficiency. Specifically, EEVEE improves average multi-benchmark scores by 10.38 and 24.32 points over Qwen3-4B-Instruct and DeepSeek-V3.2, surpassing SOTA methods GEPA and ACE by up to 37.2% and 48.2%.

Data tells stories that shape society; the data journalist's job is to turn raw information into stories non-experts can trust. A high-quality news feature takes a newsroom team weeks: hunting for context, running statistics, choosing an angle, and designing visuals. Recent agents handle individual steps well: data-science agents close the analysis loop, while design agents synthesize beautiful websites. But can an agent serve as a data journalist end to end? We introduce Data Journalist Agent (Data2Story), a multi-agent framework that orchestrates specialized roles into a single virtual newsroom. Data2Story contributes two innovations. (i) Claims are evidence-grounded: an Inspector links every number, angle, and asset back to data, code, or an external reference. (ii) Articles are multimodally generative: rather than defaulting to plain text and static charts, Data2Story reasons about what readers will want to see, then deploys multimodal tools, such as interactive maps for geography and audio for music. We evaluate Data2Story on 18 articles, each paired with the originally published expert piece, along four axes: (a) human-agent angle coverage; (b) rubric evaluation with 53 participants across five dimensions; (c) computer-use agents as judges, a cost-saving proxy for how readers navigate interactive articles; and (d) verifiability, where a coding verifier re-executes statements against the data and checks claims against references. Data2Story produces competitive, evidence-traceable multimedia stories, with particular strength in transparency and auditability. Human articles retain an edge in editorial angle, creative design, and presentation. We position Data2Story as a collaborator for journalists, enabling more evidence-based, transparent, and verifiable reporting. Code and demos are available at https://data2story.github.io.

Large-scale model training increasingly relies on composing multiple parallelism strategies, such as data, pipeline, and expert parallelism, together with memory-saving optimizations like ZeRO. Deployed systems for foundation model pretraining often rely on human experts to manually design a high-level parallelism strategy then implement the corresponding low-level execution strategy, making it difficult to adapt the system to new strategies. Meanwhile, many general-purpose frameworks are more flexible but their implementations are still tied to a fixed set of common parallelism strategies, making it challenging to integrate state-of-the-art strategies. We present Piper, a user-controllable distributed training system that decouples the strategy from the runtime implementation. Piper allows users to declare a comprehensive distributed training strategy with a small set of model annotations and scheduling directives. Each directive applies a transformation on Piper's intermediate representation (IR), a unified global training DAG that represents all computation and communication. Using this IR, Piper compiles per-device execution plans and executes them with a distributed runtime agnostic to the strategy. We show that the combined system maintains performance parity on commonly available strategies such as ZeRO, while also enabling additional performance and memory efficiency gains through joint scheduling of compute and communication in composed parallelism strategies such as DeepSeek-V3's DualPipe.

Full-duplex spoken dialogue models can listen and speak simultaneously, making them a promising architecture for natural conversation. However, current models are trained solely with supervised learning through token-level likelihood maximization, which does not directly optimize interaction-level behaviors, causing interactivity issues such as excessive silence and ill-timed turn-taking. Recent work has applied reinforcement learning (RL) to improve interactivity, but existing methods address only a limited set of interactive behaviors in their rewards. In this work, we propose a post-training alignment method that comprehensively improves the interactivity of full-duplex spoken dialogue models through RL. We address the four canonical axes of interactivity: pause handling, turn-taking, backchanneling, and user interruption. For each axis, we extract short audio segments from human conversation corpora and optimize the model with axis-specific reward functions. An extra LLM-based reward for response quality prevents semantic degradation. We apply our method to two open-source models, Moshi and PersonaPlex, demonstrating consistent improvements in interactivity on both offline evaluation with pre-recorded audio and real-time multi-turn dialogue evaluation.

Large Language Models (LLMs) are increasingly described as performing at the level of human experts on knowledge economy tasks. These claims are primarily based on how LLMs perform on benchmarking tasks that measure average performance across standardized datasets. Primary limitations of many benchmarking tasks are that they often measure performance based on content directly included in LLM training data, and they frequently do not assess the reliability of LLM performance or the magnitude of LLM errors. However, in high stakes contexts, these qualities are critically important. Through a novel LLM benchmarking task that requires writing computer code to complete a data analysis task, we compare the performance of a frontier LLM against submissions from human experts and explicitly measure the variance of responses and the magnitude of errors. Our study reveals that the human experts perform better on average on a range of metrics and demonstrate less variability in performance. Our results provide evidence that LLMs do not consistently perform at the level of human experts and demonstrate the importance of measuring variance and assessing error magnitude in LLM benchmark evaluations.

Long chain-of-thought (CoT) trajectories in large language model (LLM) reasoning cause severe inference bottlenecks due to rapid key-value (KV) cache growth. Current decoding-time compression methods mitigate this issue via token eviction, but typically assume a uniform budget distribution across all layers and heads. In contrast, existing non-uniform budget allocation methods are predominantly designed for the static prompt prefill phase, and they do not capture the stepwise context demands of autoregressive reasoning. To bridge this gap, we propose ReasonAlloc, a training-free framework that recasts decoding-time KV compression as a hierarchical budget allocation problem. ReasonAlloc operates at two complementary levels: an offline layer-wise preallocation strategy captures an architecture-driven demand pattern which we call ``\textit{Reasoning Wave}'', while an online head-wise strategy reallocates resources during decoding to information-rich heads based on real-time utility. Evaluations on mathematical reasoning benchmarks (MATH-500, AIME~2024) using DeepSeek-R1-Distill-Llama-8B, DeepSeek-R1-Distill-Qwen-14B, and AceReason-14B show that ReasonAlloc outperforms uniform-budget R-KV, SnapKV, and Pyramid-RKV (a baseline enforcing a static, monotonically decreasing layer budget), with the largest gains at small budgets (128-512 tokens). ReasonAlloc is plug-and-play with existing token-eviction policies and introduces negligible inference-time overhead.

In this work, we present COGENT, a continuous graph emulator with Neural Ordinary Differential Equations for long-term physical forecasting on irregular geospatial meshes. COGENT encodes a finite history of system states and associated forcing fields and external forcings with a graph-based history encoder, producing node-wise context vectors that capture both local spatial interactions and temporal evolution. These context vectors initialize and condition a latent Neural Ordinary Differential Equation whose dynamics are driven by interpolated future forcings and explicit relative rollout time. By modeling the forecast trajectory as a continuous latent dynamical system, COGENT can generate predictions at arbitrary future times rather than being restricted to a fixed temporal discretization. A residual decoder maps the resulting latent trajectories back to future physical states, enabling direct multi-step forecasting without repeatedly feeding predicted states back into the model. This formulation combines graph-based spatial representation, history-conditioned latent dynamics, and continuous-time rollout in a unified framework for mesh-based physical simulation emulation. In order to stabilize training with long-horizon supervision, we also propose effective rollout-horizon sampling and a progressive rollout-horizon scheduling strategy. We evaluate COGENT on transient ice-sheet simulations generated by the Ice-sheet and Sea-level System Model, demonstrating improved long-range stability over autoregressive graph baselines. These results suggest that continuous graph Neural ODEs provide a promising methodology for scalable physical forecasting on irregular geospatial meshes, particularly in applications that require stable long-horizon predictions and the ability to query system states at arbitrary times.

Recent one-step generative models accelerate sampling by learning deterministic flow maps of the underlying dynamics. These methods rely on learning from ordinary differential equations, leaving open how to define an exact distillation procedure for stochastic dynamics. We introduce the Itô map, an any-step stochastic flow map that takes an intermediate state and Brownian path and predicts future states in a single pass. The Itô map formulation yields novel estimators for inference-time control by providing cheap, differentiable access to posterior samples. Empirically, Itô maps produce diverse, conditionally valid endpoint samples from fixed intermediate states and support strong steering performance on synthetic and image-generation benchmarks. These results establish any-step SDE integration as a useful primitive for posterior sampling and stochastic control.

Flow Matching models have demonstrated strong performance across a wide range of generative tasks. However, their reliance on ODE-based iterative sampling incurs substantial computational overhead in inference, which limits their applicability in real-time scenes. While distillation is a promising solution, existing approaches largely borrow from diffusion-based score matching, often failing to exploit the intrinsic geometric structure of flows and suffering from training instability, high variance, and degraded generation quality. In this paper, we propose Mean Flow Distillation (MFD), a novel distillation framework tailored for flow matching models. We theoretically demonstrate that MFD acts as a temporal low-pass filter, effectively suppressing the high-frequency optimization noise inherent in variational score distillation (VSD) while ensuring global trajectory consistency. We further prove the Mean Flow Matching Theorem, establishing that matching expected average velocities is sufficient for strict distribution alignment. Empirically, on challenging tasks of high-dimensional manifolds including 4D occupancy forecasting and text-to-image generation, MFD achieves state-of-the-art performance, enabling high-fidelity single-step generation.

Multimodal large language models can write code to produce complex programs as well as use programs to do 3D modeling, which opens up a new avenue for 3D generation powered by their priors, world knowledge and reasoning. Yet existing benchmarks rarely evaluate 3D modeling through code. Such modeling demands more than runnable code: from a text or visual specification, a model must generate a parametric 3D program that is geometrically precise, semantically aligned and assembly-consistent. We introduce P3D-Bench, a benchmark for parametric 3D generation. Unlike a 3D mesh, a parametric 3D program exposes explicit dimensions, construction operations and part relations, revealing whether a model recovers a design's structure, not just its appearance. Under a unified protocol, P3D-Bench covers three task families (Text-to-3D, Image-to-3D and Assembly-3D) and scores each output for executability, geometric fidelity, topology, text-grounded constraints, multiview semantic alignment and part-level structure. We evaluate frontier MLLMs and text-only LLMs on 400 text cases, 400 image cases and 203 annotated assemblies, with domain-specific models as reference points. Our extensive evaluation yields three findings. First, assemblies are the hardest setting, where models still fail to compose multiple parts into a coherent structure. Second, models can often recover the global shape and semantic identity of the target object, yet fail to reproduce the precise parametric geometry specified by the input. Third, part-level modeling remains weak on assemblies, where models recover neither the geometry of each part nor the right number of parts. These results position P3D-Bench as a benchmark for evaluating precise parametric geometry and part-level structure in parametric 3D generation.

Assistive mobility and manipulation platforms have received increasing attention as a means of restoring independence to individuals with disabilities. While effective for many basic activities of daily living (ADLs), a significant percentage of everyday tasks such as opening a jar, pouring a liquid, lifting a tray, or basic meal preparation, is fundamentally bimanual and remains out of reach for any single-arm system. Adding a second arm to a wheelchair is impractical, due to the additional power draw, cost, and the loss of space required for transfers and mobility. We instead propose a heterogeneous, on-demand bimanual system, in which a wheelchair-mounted anchor arm is joined when needed by a summoned mobile manipulator that serves as a complement arm. The central technical problem, which we call bimanual joining, is conditional: the anchor has already committed to a grasp, and the complement arm must choose where to stand and what to grasp to complete the task. We formulate bimanual joining as a three-phase decomposition (plan, drive, grasp) and show that a vision-language model (VLM), coupled with standard geometric tools, provides task-level knowledge sufficient to solve a representative class of bimanual ADLs. Our system JOIN, contributes (i) a wheelchair-referenced opposition score, and (ii) task-conditioned directional manipulability. We evaluate JOIN on a Kinova Gen3 anchor and a Hello Robot Stretch~3 complement on representative same-object and different-object tasks. JOIN accomplished more attempts (19/20) than state-of-the-art methods (14/20) and required markedly less correction by the operator.

Large language models (LLMs) are rapidly acquiring capabilities relevant to biological research, from literature synthesis to interpretation of experimental data. Increasingly, LLM agents can also perform in silico biology tasks that previously required experienced human biologists. These emerging AI capabilities offer new opportunities for scientific discovery and biomedical advances, but they also shift the landscape of biosecurity risks. To address this, we introduce the Agentic Bio-Capabilities Benchmark (ABC-Bench), a suite of tasks to measure agentic biosecurity-relevant capabilities. ABC-Bench evaluates LLM agents on both benign and dual-use biology tasks: writing code to operate liquid handling robots, designing DNA fragments for in vitro assembly, and evading DNA synthesis screening. These tasks require a combination of biology and software expertise. All tested LLM agents outperformed the median expert human baseliner on all three tasks. Agents performed highly on tasks drawing on published knowledge and well-documented protocols, and more weakly on a task requiring novel bioinformatics reasoning. In three wet-lab validation experiments, we found that OpenAI's o4-mini-high produced scripts that, when run on an OpenTrons liquid handling robot, successfully assembled DNA with expected sequences.

Resistance to first-line osimertinib in EGFR-mutant non-small-cell lung cancer (NSCLC) is the canonical example of predictable clonal evolution under therapeutic pressure, yet no public benchmark exists for training or evaluating computational models on the corresponding longitudinal patient trajectories. We introduce OncoTraj, a public benchmark of 813 EGFR-mutant NSCLC patients receiving first-line osimertinib, harmonized from three real-world clinical-genomic sources: MSK-CHORD (672 patients), AACR Project GENIE BPC NSCLC (34 patients), and the FLAURA molecular-resistance supplement (107 patients). OncoTraj defines three locked tasks: (A) binary classification of progression by a fixed 12-month landmark, (B) regression of time-to-first-progression in days, and (C) six-class classification of the dominant resistance mechanism. We release the harmonized dataset, patient-level train/validation/test splits with an audited no-leakage guarantee, an open-source evaluation harness, and six reference baselines spanning a majority-class predictor, logistic regression, random forest, XGBoost, an LSTM, and a multi-task transformer. With v1's single-timepoint snapshot features, no task clears chance on clean within-source evaluation: the uniformity of this ceiling across every model class localizes the limit to the input modality (single-snapshot tissue NGS rather than serial ctDNA), not the algorithm. The benchmark does recover a reproducible literature-consistent association: TP53 co-mutation raises the 12-month progression rate from 29% to 59% cohort-wide. OncoTraj establishes a reproducible, leakage-audited baseline and converts the modality limit into concrete design requirements for a serial-ctDNA-enriched v2.

We introduce WorldOlympiad, a benchmark for diagnosing video-based world models across physical faithfulness, geometric consistency, and interaction fidelity. While existing benchmarks often focus on visual quality, semantic alignment, or short-term temporal coherence, they provide limited insight into whether generated videos obey physical rules, preserve coherent 3D structure, and sustain controllable interactions over long horizons. To address this gap, WorldOlympiad decomposes world-model evaluation into three complementary dimensions. The physical track uses object segmentation and MLLM-as-judge to assess whether generated videos follow interpretable rules in mechanics, thermal phenomena, and material properties. The geometry track reconstructs generated videos with Gaussian splatting and evaluates structural consistency, cross-view coherence, and camera-trajectory alignment. The interaction track assesses whether generated rollouts follow complex action prompts and maintain smooth, coherent transitions across consecutive video chunks. WorldOlympiad further covers three major downstream scenarios, including gaming, robotics, and general real-world videos, capturing diverse challenges from interactive control and embodied manipulation to open-domain motion and camera dynamics. Together, these tracks and scenarios form a scalable and interpretable evaluation suite that exposes failure modes beyond generic video quality. Experiments on state-of-the-art models reveal substantial gaps in physical reasoning, 3D consistency, and long-horizon interaction, underscoring the need for more structured evaluation protocols for generative world models.

Synthetic post-training pipelines commonly filter generated samples with reward models or holistic LLM judges, yet two practices remain rarely examined together: whether the filtering signal is grounded in the source evidence that induced each generation, and whether rejected samples can be systematically recovered rather than permanently discarded. We present a controlled study of both questions across gate configurations, recovery strategies, and generator scales, using adversarially injected corpora to provide ground-truth failure labels. We find that exact source provenance improves faithfulness gating for stronger judges, that hallucination and reward gates reject largely disjoint sample populations making both necessary, and that an adaptive recovery pipeline combining failure diagnosis with targeted regeneration achieves higher yield, recovery rate, and injection recall than naive resampling. Downstream fine-tuning quality is driven primarily by generator scale, with filtration and recovery conditions contributing meaningfully but secondarily.

Blood pressure (BP) is a key marker for cardiovascular risk assessment and therapeutic decision-making, and Photoplethysmography (PPG) enables low-cost, wearable-friendly cuffless BP estimation. However, even with recent progress, many PPG-based models are trained with BP regression alone and may rely on amplitude-dominated shortcuts. In addition, demographic covariates that systematically modulate vascular compliance are often incorporated only via late fusion, limiting subject-specific representation learning. We propose a Transformer-based network for cuffless BP estimation from PPG signal, leveraging self-attention to capture long-range dependencies across multiple cardiac cycles. To account for subject-specific vascular differences, the model is conditioned on demographics via FiLM-style feature modulation applied through the attention and feed-forward sublayers of Transformer blocks. In addition, we add an auxiliary morphology head to guide the model to attend to BP-relevant waveform morphology associated with arterial stiffness and wave reflection. Under calibration-based evaluation protocols on the large-scale PulseDB dataset, the proposed method achieves MAE of 4.56 mmHg for systolic BP and 2.62 mmHg for diastolic BP, reducing errors by 47% and 50% compared with prior demographic-enhanced PPG baselines. The resulting lightweight, single-sensor model supports scalable and clinically grounded cuffless BP estimation in calibration-enabled deployment settings.

Backpropagation (BP) is widely viewed as biologically implausible, in part because it requires feedback weights to be the transpose of forward weights for error propagation. Interestingly, when training a network with fixed random feedback weights to circumvent this issue, learning aligns the forward weights with the feedback weights, leading the backpropagated error signal to become an approximation of the standard gradient used by BP. This process, called Feedback Alignment (FA), occurs in MLPs and very shallow CNNs but does not scale well to deeper architectures. In this work, we first investigated differences between BP and FA models, trained on CIFAR10, specifically focusing on the effective rank of the signal. We found that the FA error has a considerably lower rank and hence is constrained to a lower-dimensional subspace compared to BP, limiting exploration of the parameter space. Motivated by this observation, we evaluated two mechanisms for increasing the effective dimensionality of FA: Muon, an optimiser that orthogonalises weight updates; and hidden activity normalisation, which promotes activation orthogonality. Across larger architectures and benchmarks, we find that these methods consistently improve over FA baselines, for example, on CIFAR100 with a Resnet-18, accuracy increases by 9 percentage points. Our results identify low-dimensional gradient dynamics as a key obstacle to scaling FA and suggest that inducing higher-dimensional update geometry is a promising route toward scaling alternatives to backpropagation.

We recast pass evaluation in football (soccer) as a Monte Carlo Tree Search (MCTS)-like evaluation problem whose components mostly exist in the literature under different names: a value model (possession value), a world model (multi-agent trajectories with ball interactions), and a policy over counterfactual actions (sampling pass variants with noise). Building on the first public high-fidelity tracking dataset with 3D ball trajectories from the Bundesliga, we introduce Monte Carlo Pass Search (MCPS), which infers kick parameters for each observed pass, samples execution variants and option variants, rolls each candidate forward with a ball-conditioned world model until the next ball interaction, and scores outcomes with a learned value model to obtain a distribution over gained value. This distribution enables distribution-aware attribution with two complementary execution-surplus scores used for analysis and ranking: mean-based and percentile-based scores. To make the world model sample-efficient under limited public data, we adapt a discrete-token, autoregressive trajectory generator from autonomous driving (SMART) and show it yields strong best-of-20 forecasting accuracy compared to baselines, while supporting fully hypothetical rollouts for downstream evaluation. We have released model checkpoints and code.

Reinforcement learning with verifiable rewards (RLVR) is a promising approach for enhancing reasoning and agentic behavior in large language models. However, rollout-intensive policy optimization is often limited by insufficient reward contrast, arising when overly simple or complex prompts generate low-variance feedback and when outcome-only rewards assign the same terminal assessment to every decision in a multi-turn rollout. Past efforts have focused on allocating available rollout resources to promising prompts, yet they only leverage sample informativeness at the prompt level and neglect variation in prefix-level informativeness across turns within the same rollout. This work targets multi-turn agentic RL by modeling each ReAct-style thought-action-observation turn as a semantically distinct node, allowing budget allocation to extend from prompt roots to turn-level prefixes with further continuations, which naturally forms tree-structured rollouts. We introduce Tree Rollout Allocation for Contrastive Exploration (TRACE), a unified rollout allocation framework that enhances reward contrast within a fixed sampling budget. Technically, TRACE allocates rollout budget to both prompt roots and intermediate prefixes that are most likely to yield mixed terminal rewards. A shared generalizable predictor estimates conditional success probability at these anchors from prefix histories to guide this allocation. The resulting adaptive tree structure enriches outcome-only feedback and amplifies the policy-update signal. Empirically, TRACE achieves competitive performance and efficiency gains on typical agentic benchmarks, e.g., improving Qwen3-14B Multi-Hop QA average accuracy by 2.8 points over competitive baselines at equal sampling cost.

We study a dynamic assortment problem on a two-sided service platform with incomplete information and heterogeneous customers in a discrete-time setting. In each period, a customer arrives seeking service, and the platform chooses an assortment of sellers to display. The customer then proposes a transaction to at most one seller in the assortment according to a multinomial logit choice model. After a fixed number of periods, sellers review the proposals they have received and each chooses at most one customer according to another multinomial logit choice model, after which the cycle repeats. A key challenge is that the platform does not know the choice-model parameters of either customers or sellers in advance. To our knowledge, this is the first study of a dynamic assortment problem in which both sides' choice parameters are unknown. We develop a data-driven algorithm that learns these parameters while optimizing the platform's objective over time. We evaluate performance using regret, which measures revenue loss relative to a clairvoyant benchmark that knows all parameters and customer arrivals in advance. We show that the algorithm's worst-case regret grows polylogarithmically over time, and we derive a matching lower bound, establishing its rate optimality.

Designing FPGA-based accelerators for modern artificial intelligence workloads requires exploring a large and complex hardware design space that involves architectural parameters, data flow strategies, and memory hierarchies, making the process very time consuming. While existing methodologies such as SECDA enable rapid hardware-software co-design through SystemC simulation and FPGA execution, identifying efficient accelerator configurations remains a largely manual process requiring extensive domain knowledge. SECDA-DSE is a framework that integrates Large Language Models (LLMs) into the SECDA ecosystem to guide design space exploration (DSE) of FPGA-based accelerators. It combines a structured DSE Explorer for generating candidate architectures with an LLM Stack that performs reasoning-guided exploration using retrieval-augmented generation and chain-of-thought prompting, coupled with a feedback loop for iterative and reinforced refinement. Building on our previous work introducing SECDA-DSE, this paper extends its evaluation by generating three accelerator designs, including element-wise vector multiplication, 2D convolution, and matrix transpose, and performing end-to-end execution on FPGA hardware. The results show that SECDA-DSE can generate SECDA-compliant accelerator designs that are successfully synthesized and executed on FPGA hardware. Furthermore, the generated designs capture kernel-specific trade-offs between compute parallelism and data movement, highlighting the potential of LLM-guided exploration to adapt architectural configurations across diverse workloads while reducing exploration time and the need for extensive human expertise.

Falls are one of the leading causes of injury and hospitalization among elderly individuals, making reliable fall awareness an essential capability for safety monitoring in residential environments. However, existing fall detection systems often rely on wearable devices or fixed sensing installations, which may suffer from low user compliance, limited spatial coverage, or degraded performance under occlusion and poor lighting conditions. In this work, we propose \textbf{EM-Fall}, an embodied fall detection framework deployed on a mobile humanoid robot. The system integrates millimeter-wave (mmWave) sensing with robotic mobility, allowing the robot to actively adjust its sensing viewpoint and maintain target observability across rooms and under occlusion. To address interference in complex residential environments, including pet motion and multipath artifacts, we design a human-centered perception pipeline combined with lightweight temporal modeling to capture motion evolution before, during, and after fall events. We evaluate the proposed system across eight real indoor environments with four participants and construct an in-home mmWave fall detection dataset. Experimental results show that the embodied mobile sensing paradigm improves monitoring continuity and maintains robust fall detection performance under diverse environmental conditions. The proposed framework provides a practical solution for robot-assisted safety monitoring in home environments.

A global shortage of trained sonographers limits prenatal ultrasound screening in low- and middle-income countries, where over half of pregnant women receive no skilled sonography. Current deep learning approaches address detection, segmentation, or classification in isolation, each demanding a separate model and expert-specified labels at inference. We present FADA, a unified vision-language model built on Qwen3.5-VL that performs clinical interpretation, classification, detection, and segmentation through a single interpretation-first pipeline without external labels. FADA distills knowledge from four domain-specific foundation models (FetalCLIP, UltraSAM, USF-MAE, UltraFedFM) via offline pre-computed feature caching. Selective distillation, which applies feature alignment only to annotation tasks while interpretation relies on standard fine-tuning, consistently outperforms full distillation across most evaluation axes. The recommended variant, FADA-SKD, achieves 0.8820 mean Dice for segmentation, 0.7671 mAP@0.50 for detection, and 100% structured interpretation compliance. Expert sonographer validation across 237 images confirms clinically acceptable outputs in both autonomous and human-in-the-loop modes, with 73.5% of interpretations scoring perfectly under clinician guidance. The system is trainable on a single consumer GPU and deployable without cloud connectivity. We validate edge deployment by running the compressed 0.8B model on a commodity smartphone (Qualcomm Snapdragon 7 Gen 1, 12 GB RAM) using llama.cpp with GGUF quantization, completing the full 5-phase pipeline in approximately 60 seconds entirely offline. This establishes a practical pathway for integrating AI-assisted fetal assessment with portable ultrasound devices, directly addressing diagnostic access gaps in resource-constrained settings. Code, models, and data are available at https://github.com/mahmoodphd/FADA.

Hallucinations, where language models (LMs) generate factually ungrounded responses, pose serious risks, as users tend to blindly rely on them. This is particularly concerning in high-stakes domains, where consequences of such model behavior can lead to significant harms. Despite notable progress in understanding hallucinations, it remains unclear how reliably these models can recognize the limits of their knowledge. We introduce PhantomBench, the first large-scale benchmark of its kind, comprising more than 60K non-existent terms and entities derived from real concepts across diverse domains. Using our benchmark, we evaluate a total of 21 models of various types and sizes. We show staggering hallucination rates across the board (with average rates as high as 86.7% in some cases), and note that even frontier models surprisingly fail to abstain on non-existent concepts, especially when the input presumes their existence. We then show that PhantomBench can serve as a proxy for studying model behavior on rare concepts for which models are more prone to hallucinate. We also provide a pipeline to construct PhantomBench, enabling scalable generation of non-existent concepts tailored to the specific needs of researchers and practitioners.

Recent deep learning approaches for network intrusion detection increasingly incorporate temporal architectures such as recurrent networks and Transformers, often reporting near-perfect performance on CIC-IDS2017. However, many existing studies neither supply their temporal modules with genuine sequence inputs nor evaluate under realistic, leakage-free conditions, making it unclear whether reported gains arise from true sequence-modeling capability. In this work, we reformulate CIC-IDS2017 as a temporal intrusion-detection task by constructing ordered flow sequences from network conversations and benchmarking nine classical and deep learning architectures under a random split, two leakage-free splits, and a padding-scheme ablation. The central finding is that padding convention, not architecture, determines the Transformer's performance: on genuinely sequential (non-padded) windows the Transformer achieves the highest macro-F1 of any model in the experiment (0.89); under zero-pad+mask evaluation it drops markedly (-0.24 macro-F1), while LSTM, GRU, and 1D-CNN remain stable. Under leakage-free group evaluation the Random Forest is the most robust model (+0.009), while the Transformer's false-alarm rate grows from 0.04% to 2.7%, a 67-fold increase invisible under conventional protocols. These findings demonstrate that evaluation methodology -- specifically padding convention and split protocol -- has a larger effect on reported performance than architectural choice, and that widely used random splits with repeat-last padding can overestimate model robustness by up to 0.24 macro-F1. We advocate leakage-free splits, explicit padding disclosure, and sequence-aware benchmarking as standard practice in future IDS research. Code and implementation details are available at https://github.com/zachmocz/temporal-ids-bench.

Elite humanoid soccer shooting requires whole-body stability, high-impulse whole-body interactions, and accuracy to targets. Motion tracking-driven reinforcement learning (RL) provides stability in whole-body movement coordination, but a fixed reference makes it hard to adapt to varied ball positions and strike timings; in contrast, task reward-driven RL struggles to explore and discover valid kicks from scratch. We therefore introduce RoboNaldo, a three-stage motion-guided curriculum RL framework for high-impulse humanoid interaction. A single human-kick reference is used as a scaffold and progressively shifts optimization towards shooting performance. The curriculum first learns a stable whole-body kicking prior, then adapts the kick to free-kick settings where the ball is stationary at random positions, and finally extends it to moving-ball shooting through a locomotion-command and kick-trigger interface. A high-level heuristic planner controls this interface during training, while alternative high-level controllers can drive the same low-level policy at inference. In simulation, RoboNaldo demonstrates free-kick shot error 48.6% lower and shoot velocity 2.96x than prior work baselines. In real world on a Unitree G1 with onboard perception, RoboNaldo attains 0.73 m and 0.86 m average target shooting error from 3 m away in free-kick and moving-ball cases, accordingly. And the post-contact ball velocity reaches 13.10 m/s, which is 59-71% of reported professional open-play shot speed. Project page: $\href{https://opendrivelab.com/RoboNaldo}{\text{opendrivelab.com/RoboNaldo}}$.

Expressive continuous control policies, such as diffusion and flow models, form the backbone of recent advances in scaling imitation learning for simulated and real robot control. While they are known to scale stably in the supervised imitation learning setting, incorporating them into reinforcement learning (RL) pipelines for policy improvement has proven more difficult. It often requires specialized training objectives or backpropagating through denoising processes, which cause well-known issues with stability and affect scalability. In this paper we study the question of whether simple policy improvement schemes at test time alone, leaving stable supervised policy training intact, can be a competitive alternative which sidesteps these issues. To this end, we propose QGF (Q-Guided Flow), an RL algorithm that performs policy optimization entirely at test time. QGF works by pre-training both a reference flow policy (via a standard behavioral cloning objective) and a value function critic and, at test time, using the value gradient to guide the reference policy to generate higher-value actions without any additional policy learning. Empirically, QGF outperforms prior test-time RL methods on single-task and goal-conditioned offline RL benchmarks with high-dimensional action spaces, and is competitive with state-of-the-art training-time algorithms while being much cheaper to run. Moreover, it exhibits favorable scaling with model size by avoiding the instability of actor-critic training, offering a practical and effective alternative RL algorithm with expressive policies.

This study investigates cross-lingual distributional skew (the Shibboleth Effect) in frontier large language models (LLMs) subjected to sustained adversarial conditions. We develop a multi-agent geopolitical wargame, the Cerulean Sea Crisis, a synthetic maritime territorial dispute designed to mirror the structural dynamics of Eastern Mediterranean conflicts. Six frontier models (GPT-4o, Llama-4, Mistral-Large, Gemini-3.1-Pro, Qwen3.6-Plus, and DeepSeek-R1) participate in a between-groups experiment (N = 10 games per arm, K = 5 rounds per game) in which the sole manipulation is the language of play (English versus Turkish), producing 586 validated statements. A zero-shot classifier assesses behavioral dispositions along two continuous dimensions: Concession Rate and Coercive Rhetoric. The results are heterogeneous. Llama-4 shows a substantial, Holm-corrected increase in coercive rhetoric under Turkish (delta = +0.800, p = .002), whereas Gemini-3.1-Pro displays an equally large decrease (delta = -0.750, p = .005). DeepSeek-R1 exhibits a similar negative shift (delta = -0.860, p = .006) and provides chain-of-thought evidence consistent with a buffering mechanism. GPT-4o shows no detectable effect (delta = +0.130, p = .614). These findings indicate that cross-lingual behavioral skew is contingent on model architecture and training regime rather than a universal property of Western-origin LLMs. We identify two distinct buffering mechanisms, chain-of-thought institutional anchoring and multilingual RLHF alignment, and discuss their implications for integrating LLMs safely into diplomatic and crisis-management settings.

Evaluation remains a critical bottleneck for interactive agent development. Existing evaluation methods often rely on static benchmarks, which fail to capture the dynamic, multi-step nature of agentic behavior and struggle to expose meaningful failure modes. While user-simulation-based evaluation offers a promising alternative, existing simulation frameworks suffer from two major limitations. First, they provide limited mechanisms for evaluating the quality and comprehensiveness of simulated interactions, making it difficult to assess whether a simulator sufficiently explores an agent's capabilities and failure modes. Second, most frameworks are restricted to either UI-only actions or API-only actions, limiting their ability to model the full range of realistic user behaviors. To address these limitations, we propose VISTA, a Versatile Interactive user Simulation Toolkit for Agent evaluation. Our toolkit includes a suite of six metrics for measuring the realism, capability coverage, and interaction effectiveness of simulated interactions. In addition, we develop a hybrid user simulator that integrates both UI-based interactions and API-based interactions, enabling more realistic and comprehensive evaluation across diverse interactive environments. We evaluate VISTA in e-commerce shopping and education customer service settings and demonstrate that it produces more realistic and comprehensive evaluations than existing methods.

Various test-time interventions for Computer Use Agents (CUAs), including critic models, have been developed to improve performance through pre-execution action evaluation in complex Graphical User Interface (GUI) environments. However, existing critics suffer from two key limitations: they (1) focus primarily on short-sighted decision loops (e.g., forgetting earlier actions) and (2) lack the visual grounding needed to detect flawed actions (e.g., clicking wrong UI elements). To address these, we introduce HiViG, a History-aware Visually Grounded test-time framework, built around a multimodal critic trained on real GUI trajectories to abstract past interactions into a compact record and to evaluate actions with visual grounding. At test time, HiViG integrates the critic into the policy decision loop to provide macro-action history, which summarizes the policy's completed achievements, and visually grounded critique, which verifies raw execution coordinates against the current screenshot to intercept errors before execution. Across web, mobile, and desktop benchmarks, HiViG consistently outperforms existing scalar and verbal critics, improving average success rates over the strongest baseline by 5.8% for Qwen3-VL-32B and 9.0% for Gemini-3-Flash, and demonstrates strong cross-platform generalization. Ablations show that macro-action history mitigates short-sighted planning and visually grounded critique reduces execution errors, with both components being critical for test-time scaling in long-horizon GUI tasks.

Aligning text-to-image flow matching models with human preferences via direct reward backpropagation is sample-efficient but hampered by two well-known pathologies: activations cannot be stored across the full sampling trajectory at modern model scale, and chained Jacobian products across steps inflate the reward gradient as it travels back to early indices. Connector-based methods, such as LeapAlign, address these issues by replacing the full backward trajectory with a short pinned path, highlighting a useful decoupling between sampling and optimization. However, the quality of the resulting gradient depends on how accurately this short path approximates the full rollout, especially over long intervals. We propose FlowBP, a unified surrogate-trajectory framework that treats the backward trajectory itself as the design object. FlowBP keeps a no-gradient cached rollout for sampling, then builds a lightweight backward surrogate from cached and selectively re-forwarded velocities. This view separates four choices: the reward-model input, active set, integration weights, and bridge coupling, and recovers prior direct-gradient methods as particular settings. Within this framework, we instantiate three variants: FlowBP-Sparse uses sparse Euler reconstruction, FlowBP-Bridge adds controlled bridge coupling, and FlowBP-Lagrange raises the order of leap quadrature. All three bound memory by the active-set size and limit gradient chaining to at most one Jacobian factor. Across SD3.5-M, FLUX.1-dev, and FLUX.2-Klein-base on preference, quality, and compositional metrics, the three variants improve over direct-gradient baselines on most metrics.

With the widespread deployment of Multimodal Large Language Models (MLLMs) in social interaction, understanding and controlling their behavior under complex personality conditions is essential. This paper introduces explicit personality conditioning and establishes a systematic evaluation framework encompassing single-personality induction, multi-personality induction, and personality switching. Experiments show that personality induction improves image captioning performance but can impair performance on tasks requiring precise reasoning, such as visual question answering (VQA). Balancing and residual effects are observed during multi-trait composition and dynamic switching, indicating that model behavior is co-modulated by both previous and current personality constraints. Existing prompt-based personality induction methods show limited transferability to multimodal settings. Our work reveals the dynamic and complex nature of personality modeling in MLLMs and underscores the need for robust, tailored methods for personality induction and evaluation. The code will be released when the paper is accepted.

Recent advances in reasoning and tool-calling capabilities of large language models (LLMs) have enabled increasingly capable agentic systems. However, existing benchmarks remain limited in task complexity, realism, and domain diversity, and often fail to capture interactions that span multiple domains, limiting their ability to evaluate agents in realistic multi-step settings that require sustained reasoning and coordination. To address these limitations, we introduce T1-Bench, a high-fidelity, comprehensive benchmark for evaluating agentic systems in realistic customer-facing, multi-domain environments, featuring interleaved scenarios that require structured reasoning across multi-turn user-assistant interactions and substantially increasing both compositional complexity and evaluative rigor across 25 domains of varying difficulty. We evaluate T1-Bench using 12 proprietary and open-weight models, providing a reproducible and standardized framework for assessing agent behavior, tool utilization, and conversational quality in complex, multi-step environments. We further complement automatic evaluation with human judgments to strengthen the assessment of qualitative performance. Overall, T1-Bench substantially advances prior benchmarks by increasing task complexity, interaction depth, and domain coverage in simulated multi-domain environments. To facilitate future research on agentic systems, we will publicly release data and evaluation code as open source.

AI control protocols oversee untrusted models by monitoring their actions and modifying potentially unsafe steps, often using a trusted model. This partially tampers with the untrusted model's trajectory. If the trusted model detects such an intervention, it may infer properties of the monitor and adapt to evade control. We introduce \textbf{CIAware-Bench}, a benchmark for measuring \textbf{c}ontrol \textbf{i}ntervention (CI) awareness across frontier models. CIAware-Bench tests whether models can distinguish their own trajectories from those modified by a control intervention. The benchmark is comprised of a suite of four task domains (essay writing, BigCodeBench, Bash Arena, and SHADE-Arena), while varying trajectory watermarking, side-task presence, and the control protocol. Evaluating eleven frontier models, we find low to moderate CI awareness under default settings (up to 0.87; random chance balanced binary classification accuracy is 0.5) with substantial variation across task domains and model pairs. Detection is generally easier across model families, suggesting that models exploit provider-specific differences in style or post-training. Overall, CI awareness is not a fixed model-level property, and should be measured for each new model release and deployment scenario. We release CIAware-Bench to track CI awareness and inform control protocols whose interventions are harder to detect.

Chain-of-thought (CoT) supervised fine-tuning (SFT) is widely adopted to improve reasoning ability, yet we find that it systematically degrades long-context recall in hybrid linear-attention models. Across architectures including HypeNet and Jet-Nemotron, retrieval performance on Needle-In-A-Haystack (NIAH) deteriorates substantially after CoT-SFT, and the degradation becomes more severe under harder retrieval settings and longer context windows. For example, HypeNet-9B on NIAH-S2@256K decreases from $67.2\%$ to $9.4\%$. We attribute this to CoT-SFT biasing attention gradients toward short-range patterns, disrupting query-key projections ($W_Q, W_K$) that are responsible for long-range routing. Motivated by this observation, we propose QK-Restore, a training-free method that restores only $W_Q$ and $W_K$ from the pre-SFT checkpoint while preserving all other post-SFT parameters. We further introduce a Procrustes variant to balance routing preservation and reasoning adaptation. Across architectures, QK-Restore consistently restores long-context capability at zero training cost while preserving reasoning performance; for instance, on HypeNet-5B it improves S3@256K from $65.4\%$ to $76.4\%$ while maintaining strong reasoning performance.

Effective demand response in smart microgrids requires prosumers to cooperate voluntarily under strategic self-interest, a coordination problem structurally equivalent to a repeated Prisoner's Dilemma on a social network. This paper presents a multi-agent simulation in which a Large Language Model (LLM) Influence Compiler issues structured demand-response directives to a population of heterogeneous prosumer agents, each governed by a hybrid decision architecture combining game-theoretic base probability (derived from payoff history, neighbour imitation, and exploitation memory) with LLM narrative evaluation of incoming coordination signals. The hybrid architecture resolves a key methodological challenge: LLMs aligned via Reinforcement Learning from Human Feedback (RLHF) exhibit strong cooperation bias when used as direct decision-makers, producing flat dynamics regardless of grid conditions. By separating strategic reasoning from grounded narrative evaluation, the model generates realistic prosumer behaviour across six personality archetypes, with baseline cooperation near 50% and clear differentiation under influence. Compiled structured directives achieve 33.3% demand-curtailment cooperation versus 27.0% for unstructured messaging and 28.0% for a no-intervention baseline ($Δ_\mathrm{comp} = +0.063$), with the advantage preserved across both grounded and idealized agent substrates ($Δ= +0.083$) and across all resistance levels ($R = 0.1$ to $0.7$). Hub-targeted dissemination via high-centrality network nodes outperforms peripheral or random targeting, confirming that grid topology provides mechanistic amplification independent of message content. These results suggest that structured LLM compilation, grounded agent reasoning, and network-aware targeting are complementary design principles for scalable, interpretable demand-response coordination in smart-city energy systems.

Instruction-tuned LLMs are increasingly converted into reasoning models through post-training to improve multi-step task performance. This conversion is usually optimized for reasoning accuracy, without explicitly preserving the alignment behavior of the instruction-tuned model, such as safe refusal, bias avoidance, and privacy protection. We ask: does this conversion preserve alignment? We study this question through a trustworthiness audit and find that it is not behavior-preserving by default. For a systematic analysis, we compare reasoning models produced via supervised fine-tuning, RL-based post-training, and distillation against matched instruction-tuned baselines across six trustworthiness dimensions: safety, toxicity, stereotyping and bias, machine ethics, privacy, and out-of-distribution robustness. We observe that reasoning models often improve on reasoning benchmarks but exhibit alignment regressions, including increased toxicity, amplified stereotyping, miscalibrated refusal, and contextual privacy leakage. These regressions are consistent with behavioral drift from the instruction-tuned baseline, measured by KL divergence. Overall, our results point to the broader conclusion that trustworthiness metrics are essential for evaluating reasoning models and should be reported alongside gains in reasoning capability.

Reusing a held-out benchmark adaptively should, in principle, invite overfitting. Yet benchmark-driven machine learning (ML) has produced surprisingly little overfitting in practice. An attractive hypothesis is that successful ML strategies are highly compressible. We study this in the setting of LLM-driven research agents, where the hypothesis becomes directly testable via two complementary information bottlenecks. In \emph{output compression}, an exploration agent adaptively searches for high-performance models using a validation set, and we test whether a fresh ``reproducer agent'' can reproduce its performance given only an extremely short prompt and the training data. In \emph{input compression}, the explorer receives only one-bit feedback indicating whether each submitted model improves on the running best. Across 8 datasets spanning tabular classification, vision, language modeling, diffusion modeling, and reward modeling, we find that these bottlenecks have little effect on performance: short prompts and compressible feedback are sufficient to reproduce and find high-performance models. The hypothesis is falsifiable: when we deliberately induce validation-set overfitting, the results fail to reproduce with short prompts. Taken together, our results support a description-length explanation for the lack of overfitting in benchmark-driven ML: successful strategies occupy a low-complexity region of strategy space.

Recent years have witnessed the rapid evolution of AI agents toward handling increasingly complex, real-world tasks. However, existing benchmarks rarely evaluate whether agents can operate graphical user interfaces to complete long-horizon, high-value professional workflows across diverse domains. Current GUI benchmarks still predominantly focus on general-purpose software, relatively simple applications, and short-horizon tasks, leaving it largely unknown whether modern agents can follow user instructions to autonomously operate domain-specific professional software and accomplish economically valuable work in an end-to-end manner. To bridge this gap, we introduce Workflow-GYM, a benchmark for long-horizon GUI tasks centered on professional domains and specialized software environments. Through extensive experiments on state-of-the-art models, we find that even the strongest models achieve only slightly above 30% success rates, highlighting that professional long-horizon GUI workflows remain highly challenging for current GUI agents. Further analysis reveals that current agents struggle to maintain long-horizon workflow consistency, frequently exhibiting workflow stage omission, error propagation, objective drift, and insufficient understanding of professional software environments. Our findings provide important insights into the limitations of current agent systems and suggest key directions for the next generation of GUI-agent research.

Spiking Neural Networks (SNNs) coupled with neuromorphic hardware offer energy-efficient solutions for humanoid robot control. However, existing SNN-based motor control systems address bipedal locomotion and arm control in isolation, leaving integrated control of both unaddressed. We present a spiking architecture that coordinates force-based arm control and bipedal locomotion in a simulated humanoid, using the Neural Engineering Framework (NEF) and Semantic Pointer Architecture (SPA). High-level action selection between locomotor and arm control is mediated by a biologically grounded spiking basal ganglia model. We validate the system through co-simulation of Nengo, for the neural control, and Isaac Sim, demonstrating successful target reaching, continuous digit drawing, path-following locomotion, and finally, switching between walking and arm control via basal ganglia disinhibition. To our knowledge, this is the first integrated spiking controller to combine bipedal locomotion and arm control on a full-scale humanoid platform. The full spike-based implementation enables future deployment on low-power neuromorphic hardware.

Recent efforts to extend large language models (LLMs) to speech inputs typically rely on cascaded ASR-LLM pipelines, end-to-end speech-language models, or bridge/distillation-based adaptation. While these routes respectively reuse strong pretrained components, enable native speech-language interaction, or offer lightweight adaptation, they often suffer from transcript-interface latency, costly multimodal training, or sequential speech-language coupling. To address these limitations, we present AuRA, a method that distills audio encoding capability into the LLM. Specifically, AuRA feeds the same speech input to an ASR encoder (as a teacher) and a LoRA-adapted LLM (as a student) through a lightweight audio embedding layer, and uses layer-wise distillation to align the student's hidden states with corresponding teacher representations, thereby internalizing speech representations into lightweight LLM-side adaptations. Compared with cascaded and serial bridge methods, AuRA enables tighter speech-language joint modeling and efficient parallel end-to-end inference, while also reusing pretrained speech and language models rather than requiring large-scale multimodal training. On multiple speech-language benchmarks, AuRA consistently outperforms cascaded systems, speech-to-LLM adaptation baselines, and large-scale speech-language and multimodal models in both effectiveness and efficiency.

Airport surface operations increasingly constrain performance at high-throughput hubs. This study examines arrival taxi-in decisions at Hartsfield-Jackson Atlanta International Airport (KATL) and proposes a two-stage, data-driven decision aid that mirrors controller workflow. Stage I predicts the runway exit selected by an arriving aircraft. Stage II predicts whether, given that exit, the aircraft will cross the active departure runway at a designated point or use the end-around taxiway. Models are trained using ASDE-X surface trajectories, aircraft characteristics, ramp destinations, short-horizon traffic rates, and weather across multiple look-back windows. We benchmark nine classifiers, including Random Forest, XGBoost, LightGBM, and CatBoost, and evaluate accuracy, macro-F1, precision-recall behavior, confusion matrices, Brier score, and Expected Calibration Error. Across east and west flows, XGBoost and LightGBM outperform Random Forest. Stage I achieves 0.86-0.89 accuracy with macro-F1 scores of 0.40-0.50, while Stage II achieves 0.70-0.74 accuracy with macro-F1 scores of 0.28-0.55. Feature-importance analysis shows that approach speed is the main driver of exit choice. Departure rate, crossing rate, ramp destination, and, for west flow, the selected exit are the strongest predictors of crossing versus end-around routing. Minority classes remain harder to predict because of feature-space overlap, as shown by t-SNE and UMAP analyses. The proposed framework supports controller situational awareness through calibrated, explainable predictions while preserving human responsibility for final routing decisions.

Tuning controllers for strongly coupled multi-input multi-output (MIMO) industrial processes is hard: decentralized classical auto-tuning ignores loop interaction, and local numerical optimization from natural initializations stalls in the resulting non-convex cost landscape. We ask whether on-premise open-source large language models (LLMs), which keep data on-site and need no plant model, can help. On a single-loop CSTR, classical relay-feedback tuning (IAE 0.106, near the 0.102 optimum) beats an LLM tuner (0.162): for simple loops the LLM adds nothing. The picture inverts on a strongly coupled quadruple-tank with conflicting set-points, scored by a penalized cost J = IAE + lambda*TV(u) that rewards tracking without chattering actuators. There, naive relay tuning (J ~ 28.6) and naive LLM tuning (29.7) are no better than open loop (22.7), and a local optimizer from balanced starts fails in 10/10 runs. A scaffolded open LLM instead reasons about the coupling, proposes the counter-intuitive asymmetric structure, and reaches J ~ 16.9 +/- 0.2 from any start; refining it with a classical optimizer attains the smooth global optimum (J ~ 12.0, 10/10 vs. 0/10), which even applies a non-obvious negative integral correction decentralized tuning cannot. A global optimizer (differential evolution) also reaches this optimum, so the LLM is not the only route; its advantage is sample efficiency and interpretability: a usable controller in 18 evaluations (where the global optimizer is worse than open loop) plus a stated rationale. This edge grows with dimension, reaching ~6x fewer evaluations on a 3x3 plant. The behaviour generalizes across four open models, and on a benign plant the LLM offers no advantage, sharpening the boundary. We contribute a reproducible benchmark delimiting when open LLMs help in control tuning: not as optimizers, but as a sample-efficient, interpretable structural prior.

OpenClaw has rapidly emerged as a transformative artificial intelligence (AI) agent framework, and its ability to autonomously execute complex, multi-step tasks has attracted an ever-growing and diverse user base. However, this capability comes with significant risks. While existing research has made important strides in characterizing these threats, such work is predominantly directed at technically sophisticated audiences. It remains largely inaccessible to non-technical users. This demographic now makes up an increasingly large and underserved portion of the community, yet it is these very users who most urgently need practical and straightforward guidance. In response, we bridge this gap through a series of interconnected efforts designed to lower the risk barrier for non-technical OpenClaw users. First, we identify and categorize seven core risks that OpenClaw users may encounter in daily usage, explaining each in plain language so that non-technical users can readily grasp the nature and potential consequences of these threats. Second, for each identified risk, we distill a set of corresponding defensive strategies into clear and actionable operational steps that are easy to follow. Third, to make protection even easier, we provide a companion OpenClaw Skill that automates key security configurations, enabling users to safeguard their systems with minimal manual intervention. Through this work, we demonstrate that safeguarding against the risks of intelligent agents need not be the exclusive domain of security experts, and that non-technical users can meaningfully participate in reducing these risks through simple, practical actions.

Zinc-based alloys are indispensable emerging absorbable metallic biomaterials, and their macroscopic performance is governed by microstructural characteristics. Intermediate phases-key microstructural constituents-are pivotal in regulating mechanical and functional properties. However, intermediate phase segmentation in zinc alloy microstructures faces formidable challenges: scarce annotated datasets, low contrast, difficulty detecting small targets, and heterogeneous morphologies. To this end, we construct IPSM-Bench, the largest high-quality dataset for zinc-alloy intermediate phase segmentation. Furthermore, we propose SCoP-SAM, a new Spatial Context Prior-guided SAM method that leverages the gradient structure and grayscale properties of intermediate phases to capture spatial context priors and incorporates them into the entire SAM encoding-decoding process, improving segmentation performance. Based on the proposed IPSM-Bench, we establish a new benchmark for intermediate phase segmentation to systematically evaluate state-of-the-art (SOTA) methods and advance research on zinc alloy microstructure analysis. Extensive experiments on IPSM-Bench and additional public alloy benchmarks demonstrate that our SCoP-SAM not only achieves SOTA performance for zinc-alloy intermediate phase segmentation but also generalizes remarkably well to other alloy scenarios.

Many Markov decision processes (MDPs) in operations research have feasible actions that are state dependent and defined implicitly by various operational constraints. These features make it difficult to use standard deep reinforcement learning (DRL) algorithms, whose action interfaces typically assume either a fixed finite action catalog or a simple Euclidean space. Motivated by a Taylor expansion of the optimal action-value function, we propose Bellman--Taylor score decoding, a framework that moves policy learning to a Euclidean score space while enforcing feasibility through an action decoder. The induced latent-score MDP then can be optimized by standard DRL algorithms without differentiating through the decoder. We provide a performance guarantee showing that the optimality gap of this approach decomposes into a structural approximation error and an algorithmic learning error. Lastly, we apply this framework to a queueing network control problem, where the policy essentially learns a state-dependent index-based dispatching rule. Numerical experiments show near-optimal performance in small instances and considerable improvements over benchmarks in larger systems.

The driving behavior of autonomous vehicles is typically governed by the cost function of their motion planner, which encodes objectives such as speed tracking, smoothness, lane keeping, and collision avoidance. However, tuning the parameters that shape this cost function is a challenging task that requires technical expertise, limiting the vehicle's ability to adapt to evolving traffic scenarios or end-user preferences. This work presents a language-driven framework for adaptive cost design in autonomous driving. A Large Language Model (LLM) interprets structured scenario descriptions and natural language user queries to generate the parameters applied to a risk-aware Model Predictive Path Integral (MPPI) controller. The system incorporates a human-in-the-loop validation stage in which the proposed behavioral changes are described in non-technical language and confirmed prior to deployment. Users may additionally provide feedback either before or after deployment, enabling iterative refinement of the vehicle's motion behavior. The framework is evaluated across multiple queries in realistic driving scenarios to assess its effectiveness. Simulation results demonstrate that the method successfully induces behavioral changes that align with the intended requirements in an intuitive manner, thereby bridging the gap between intelligent vehicle control systems and end users.

Visual in-context learning has been proposed as a pathway towards dynamic models that can generate predictions based on a provided context and thereby can adapt to new vision tasks at test-time. Yet, the evaluation of the adaptation capabilities of these models has been limited to narrow setups that mainly mirror tasks or image domains from pre-training for which real adaptation is not required. We address this gap by constructing a broad Visual In-Context BEnchmark (VIBE) with a focus on diverse imaging domains and a wide range of tasks. With this, we are able to get a much clearer picture of the adaptive capabilities of visual in-context models when faced with new image- and task distributions. We stress test six models on $14$ datasets and $12$ tasks (in total, we explore $106$ dataset-task combinations) and compare them under a unified, reproducible evaluation protocol, in an one-shot setting. Our evaluation uncovers key insights on the state of visual in-context learning, including limitations, systematic failure modes and promising directions. To foster broader evaluation, we will openly release our VIBE toolkit.

The deployment of Large Language Model (LLM) agents for computer automation is accelerating, yet their ability to navigate complex, professional-grade productivity software is largely untested. We argue that Office automation is an ideal environment for benchmarking document-automation capability, as it requires long-horizon planning and reasoning, precise parameter configuration, and multi-application integration. To quantify this capability, we introduce an evaluation based on China's National Computer Rank Examination (NCRE), featuring 200 comprehensive practical-operation tasks across Word, Excel, and PowerPoint. Each task is scored on a 100-point rubric scale using 7,118 machine-gradable criteria, and Score Rate (SR) denotes the mean percentage of rubric points earned across these tasks. We benchmark 7 frontier LLMs and observe stark limitations: single-turn models score a maximum of 36.6%. A stronger agentic system with execution feedback, iterative repair, and broader Office automation access reaches 68.8%, but remains below the 95.5% community-reference score used as a scoring sanity check. Ultimately, our experiments demonstrate that despite recent advancements in code generation, achieving reliable fine-grained Office document automation remains a significant challenge for current code-generating LLM and agent systems.

Persistent memory systems promise to make LLMs more helpful by storing user beliefs over time. We show they also make models less correct by systematically amplifying sycophancy, wherein models prioritize agreement with users over accuracy. We conduct the first systematic evaluation of this effect, introducing MIST: a benchmark of synthetically generated multi-turn conversations where users express plausible misconceptions in scientific, medical, and moral reasoning domains. Testing across three state-of-the-art memory systems and five model families reveals that memory amplifies sycophantic behavior across all conditions, with up to 25x higher sycophancy rates than in-context baselines. Error analyses suggest memory extraction as the primary culprit: lossy compression into discrete snippets encodes user misconceptions while discarding corrective context. Based on these results, we propose two lightweight mitigations that substantially reduce sycophancy while matching or exceeding memory systems at factual recall.

We introduce a new tool, Express, for converting a non-causal attention approximation into a causal approximation with matching approximation guarantees. When combined with the state-of-the-art Thinformer approximation, Express improves upon the best known causal attention guarantees, delivering $\log^{3/2}(n)/s$ approximation error with only $O(s)$ memory and $O(s^2 \log^2(n))$ compression overhead for a sequence of length $n$. We pair these developments with an efficient I/O-aware Triton implementation, demonstrate substantial speedups over FlashAttention 2, and use Express to overcome four resource bottlenecks in the language modeling pipeline: long-context prefill, KV cache compression, long-form memory-constrained decoding, and long-form compute-constrained decoding.

As artificial intelligence and machine learning (AI/ML) models become integral to network operations, their lack of transparency poses a significant barrier to operator trust. Existing explainable artificial intelligence (XAI) techniques often fail to bridge this gap for non-specialists, producing technical outputs that are difficult to translate into actionable insights. This paper presents a framework specifically designed to address this shortcoming. It leverages a moderately sized large language model (LLM) and extends beyond the standard use of SHapley Additive exPlanations (SHAP) feature influence values. The framework employs a structured prompt enriched with mutual feature interaction data to generate human-understandable natural language explanations. To validate our framework, we performed an empirical evaluation on an optical quality of transmission (QoT) estimation use case with human evaluators. We collected independent performance evaluations from specialists, which showed a high inter-evaluator agreement. Compared to a state-of-the-art baseline that uses only SHAP feature influence values in a straightforward prompt, our approach improves the explanation usefulness and scope by 12.2% and 6.2%, while achieving 97.5% correctness.

Image steganalysis, the detection of hidden information embedded in digital images, is a core component of modern cybersecurity and digital forensics. Recent residual Transformer architectures, such as the Pixel-Difference-Convolution and Enhanced-Transformer-Network (PENet) [1], achieve strong detection accuracy, but their computational and memory demands hinder deployment in resource-constrained settings. We present PENet+, a lightweight steganalysis framework that preserves PENet's discriminative structure while substantially improving efficiency. Rather than redesigning or compressing the attention blocks, we retain PENet's self-attention topology for reproducibility and add a classifier-streamlining stage that progressively narrows the SPP-to-FC1 input channels (SPP: spatial pyramid pooling; FC1: first fully connected layer), yielding large reductions in parameters and FLOPs with negligible accuracy loss. We further refine the high-pass-filter (HPF) stem with an activation-aware mechanism that aggregates HPF responses early and selects a balanced SRM-Gabor top-K subset, and we replace PENet's backbone with a MobileNetV2-style inverted residual network. A balanced configuration with K=31 filters (16 Gabor + 15 SRM) matches or surpasses heavier settings at lower compute. Finally, we motivate PReLU from a steganalysis standpoint, arguing that preserving negative responses helps capture weak stego cues that ReLU suppresses. On a disjoint ALASKA2 JPEG QF90 protocol at 512x512 resolution (5,000 cover images for training, validation, and internal testing; a separate 19,000-cover evaluation set), PENet+ achieves up to 45.5% fewer parameters and about 97% fewer FLOPs than the re-evaluated PENet baseline, offering a computationally efficient direction for resource-constrained steganalysis. Device-level latency and power measurements remain future work.

A common assumption holds that enough observational and interventional data, given to a strong enough predictor, suffices. We report a failure mode that contradicts it. Across hundreds of structural causal models, on identified quantities a strong predictor and a Bayesian baseline both succeed, but on unidentified quantities (the couplings between counterfactual worlds) the predictor collapses to a point, on 28% of models to one no valid model can produce, while the truth is an admissible interval more data never narrows. The gap is structural: prediction cannot represent uncertainty over counterfactual couplings. We cast a world model as a single positive semidefinite coupling kernel K(T,T') over admissible worlds, whose diagonal is the ordinary posterior (what a predictor recovers) and whose off-diagonal is the cross-world coupling it cannot, which every counterfactual reads. The paper is the theory of that off-diagonal. It is real: two states with identical posteriors differ on a cross-world query, and the off-diagonal is the coupling that fixes counterfactuals. It can be bounded: positive semidefiniteness is partial-identifying information the marginals lack, and enforcing it bounds counterfactuals in polynomial time where the exact response-type program is intractable. Logical structure sharpens it: ontology axioms tighten the bound by up to a third, propagating to couplings they never touch. It can be acquired: targeted scars, constraints learned from encountered infeasibilities, close the gap several times faster than untargeted ones. Its full reconstruction is approximate counting of the admissible worlds, tractable below the Sly-Sun threshold and inapproximable above; we do not claim to beat the worst case.

LLM-based coding agents are usually evaluated in familiar software settings: mainstream languages, common libraries, and public repositories. These benchmarks remain important, but they can hide how agents behave when the language itself is unfamiliar. We evaluate six contemporary coding agents on four esoteric programming languages using a sequential setup with file editing, local execution, and hidden-test grading. Our protocol exposes capability differences between these agents that mainstream coding and agentic benchmarks such as SWE-Bench Verified and Terminal-Bench 2.0 compress into much narrower bands. We observe that the strongest agents, Claude Opus 4.6 and GPT-5.4 xhigh, often avoid writing the target language directly. On Brainfuck and Befunge-98, they write Python programs that generate target-language code and debug those generators locally. Forbidding this metaprogramming strategy causes large performance drops. Text guidance distilled from this strategy does not materially improve weaker agents. In contrast, Opus-derived Python helper code for building generators, with no solved benchmark programs or hidden-test answers, sharply improves Sonnet 4.6 and GPT-5.4 mini on the same problems, while Haiku 4.5 remains low. More interpreter calls and output tokens improve stronger agents but leave weaker agents near their original performance, indicating that these resources amplify useful strategies rather than create them. Together, these results show that strong coding agents adapt to unfamiliar languages by using tools, feedback, and workspace state to build a working model of the target language. Metaprogramming is the clearest case, but the broader gap is constructing and debugging a strategy that works under the target language's rules.

Warning: This paper contains several toxic and offensive statements. Modern large language models (LLMs) are typically aligned through large-scale post-training to ensure fair and reliable behavior. In this work, we investigate how easily such guardrails can be broken by Group Relative Policy Optimization (GRPO). We show that one-shot GRPO training on a single biased example is sufficient to induce systematic bias, with stereotype-driven reasoning generalizing across attributes, categories, and benchmarks. We further find that models differ in their susceptibility based on the initial likelihood of producing biased outputs. Our results reveal a critical vulnerability in post-training: alignment can be overridden by a single example.

Large language models can reduce the manual effort required to set up finite element simulations, but they introduce reliability risks when generated solver code lies on the critical path. We present a constrained natural-language interface for multi-physics finite element analysis in which the LLM is limited to front-end tasks: parsing prompts into structured JSON, generating Gmsh code only for non-catalog geometries, and using retry feedback for those stages. It never writes FEniCS solver templates, derives weak forms, or writes the numerical solver core. A deterministic dispatcher maps the validated specification to five human-written FEniCS/UFL templates: linear elasticity, hyperelasticity, elastoplasticity, thermo-mechanical coupling, and phase-field fracture. We validate this deterministic template layer against analytical solutions and published 2D/3D benchmarks. Smooth cases reach sub-percent agreement on adequate meshes, while harder nonlinear cases reach the 2-5 percent range. We also evaluate the LLM-facing front end directly. In a 15-prompt parser benchmark, first-pass valid parses were obtained for 9 cases, and all remaining cases were repaired after retry, giving a final valid parse rate of 100.0 percent, 100.0 percent problem-class accuracy, and 97.1 percent field-extraction accuracy. In a 10-case custom-geometry benchmark routed through the real LLM-to-Gmsh path, first-pass and final success were both 90.0 percent, with one unrecovered invalid-geometry failure. These results show that the parser and constrained prompt/validation design are effective on these benchmarks. As an end-to-end demonstration, the system generates and analyzes a 3D elastoplastic L-bracket with a fillet and bolt hole from one natural-language prompt. The contribution is a measured architecture for natural-language-driven variational simulation, not open-ended autonomous code generation.

Lifelong embodied navigation in dynamic environments requires robots to form persistent scene understanding from fragmentary observations, which remains difficult for existing methods that rely on explicit maps or scene graphs and struggle to generalize beyond structured settings. We propose AllDayNav, a lifelong self-learning navigation framework that implicitly encodes scene dynamics into the billion-scale parameters of a large model via reinforcement learning, powered by a self-evolving multimodal memory that maintains and updates visual keyframes, semantic descriptions, and temporal context while autonomously generating open-vocabulary instructions, image goals, and structured rewards. Experiments in both synthetic and real-world environments across cross-room, cross-episode, and cross-task scenarios show that AllDayNav achieves success rates approaching $100\%$ and consistently surpasses strong map-based, VLM, and RL baselines in path efficiency and robustness, demonstrating implicit, memory-driven reinforcement learning as a scalable alternative to explicit mapping for reliable lifelong navigation.

Multi-terminal DC (MTDC) transmission systems based on modular multilevel converters (MMCs) are a key component of the envisioned future energy sector, where sustainability and efficiency are increasingly prioritized. To ensure their reliable operation, MMC currents must be regulated safely and rapidly under a wide range of uncertain operating conditions. Consequently, the design of current controllers faces a fundamental challenge: achieving fast transient response while maintaining robustness against uncertainties. This paper addresses this challenge by proposing a linear matrix inequality (LMI)-based design framework that leverages Lyapunov stability conditions to synthesize a less conservative static state-feedback controller. The proposed design method explicitly accounts for system constraints, including input saturation and overcurrent limits. The proposed method effectiveness is assessed on the CIGRE MT-HVDC benchmark, simulated in RTDS, and compared with existing methods.

Long-document question answering (QA) requires large language models (LLMs) to reason over evidence scattered across lengthy documents, where answers often depend on event order, section-level context, and cross-part evidence connections. Although retrieval-augmented generation (RAG) reduces the input context by retrieving relevant evidence, existing structured RAG methods still face three limitations: costly query-agnostic knowledge organization, insufficient use of original document structure, and no reuse of historical reasoning experience. To address these limitations, we propose DocTrace, a multi-agent RAG framework for long-document QA that supports query-triggered knowledge organization, document-structure-aware and experience-guided reasoning. DocTrace preserves document hierarchy with a lightweight document structural tree index, constructs agent-shared hypergraph-structured working memory on demand during reasoning, and stores successful reasoning plans in graph-structured experience memory for future reuse, enabling adaptive exploration across related long-document questions. Experiments on four long-document QA datasets show that DocTrace achieves the best performance on three datasets, surpassing the strongest baseline, ComoRAG, by up to 8.85% in F1 and 4.40% in EM, while reducing the overall computational cost by 53.32%

The recent trend in scaling models for robot learning has resulted in impressive policies that can perform various manipulation tasks and generalize to novel scenarios. However, these policies continue to struggle with following instructions, likely due to the limited linguistic and action sequence diversity in existing robotics datasets. This paper introduces Task Robustness via Re-Labelling Vision-Action Robot Data (TREAD), a scalable framework that leverages large Vision-Language Models (VLMs) to augment existing robotics datasets without additional data collection, harnessing the transferable knowledge embedded in these models. Our approach leverages a pretrained VLM through three stages: generating semantic sub-tasks from original instruction labels and initial scenes, segmenting demonstration videos conditioned on these sub-tasks, and producing diverse instructions that incorporate object properties, effectively decomposing longer demonstrations into grounded language-action pairs. We further enhance robustness by augmenting the data with linguistically diverse versions of the text goals. Evaluations on LIBERO demonstrate that policies trained on our augmented datasets exhibit improved performance on novel, unseen tasks and goals. Our results show that TREAD enhances both planning generalization through trajectory decomposition and language-conditioned policy generalization through increased linguistic diversity.

Although Large Language Model (LLM) agents have demonstrated strong performance on complex tasks, their learning is often limited by inefficient interaction feedback and static training environments, which hinder broader generalization. To address these limitations, this paper introduces Role-Agent, \textcolor{black}{a framework} that harnesses a single LLM to function concurrently as both the agent and the environment, enabling a bootstrapped co-evolution. Role-Agent comprises two synergistic components: World-In-Agent (WIA) and Agent-In-World (AIW). In WIA, the LLM acts as the agent and predicts future states after each action; the alignment between predicted and actual states is then used as a process reward, encouraging environment-aware reasoning. In AIW, the LLM analyzes failure modes from failed trajectories and retrieves tasks with similar failure patterns, thereby reshaping the training data distribution for targeted practice. Experiments on multiple benchmarks show that Role-Agent consistently improves performance, yielding an average gain of over 4\% over strong baselines.

Reconstructing local stress fields in heterogeneous microstructures under non-linear, history-dependent loading remains a major computational bottleneck in multi-scale simulations. We propose a coupled LSTM-GNN framework that links the temporal and spatial aspects of local stress field reconstruction. A Long Short-Term Memory network encodes macroscopic stress-strain sequences into a compact hidden state that captures the path-dependent constitutive response, while a physics-informed Graph Neural Network reconstructs the spatially-resolved stress field at each time step. We introduce a relative weighting strategy with linear warm-up to balance the data-driven reconstruction loss and a discrete divergence-based equilibrium penalty. This resolves the scale mismatch that prevents fixed-weight formulations from converging in the elasto-plastic regime. The model is trained on 10,000 non-proportional loading paths applied to a periodic plate-with-a-hole microstructure and von Mises elasto-plasticity. The model achieves three orders of magnitude speedup over finite element simulations and generalizes to loading sequences twice the training length, with 1.9% cumulative error. Because the graph relies on mesh connectivity instead of the specific element type, one trained surrogate can be applied directly without retraining to meshes with different element types and to both coarser and finer resolutions, while in all cases reproducing the high-fidelity quad-element FE field used during training. Indeed, the message passing characteristics inherent to GNN and MeshGraphNet architecture render the model mesh-agnostic. Analysis of the LSTM hidden states suggests a low-dimensional structure related to the internal state variables of the constitutive model.

We introduce a spoofing countermeasure architecture conditioned on speaker-reference recordings, but observe that it converges to a solution that effectively ignores the reference during inference. Surprisingly, training with a reference channel induces invariance that improves deepfake detection, even when the reference is absent or mismatched during inference. Based on this observation, we propose a Reference-Augmented Training (RAT) strategy. RAT yields improved detection performance compared to single-utterance baselines, even when the reference recording is replaced with a zero vector at inference. Through rigorous analysis, we demonstrate that the optimization process rapidly diminishes the reference contributions, leading to inference largely independent of the reference channel. Using RAT, we achieve state-of-the-art 2.57% EER and 0.074 minDCF on the ASVspoof 5 benchmark with a single detector, surpassing even large ensemble systems.

Visual in-Context Learning (VICL) aims at making progress towards adaptive vision models, that can -- based on a few examples -- adapt to a new task at test-time. With the history of in-context learning in natural language processing research, where large, parameter-heavy models are in use, one pathway that current VICL methods take is model- and data-scaling as key ingredients. Yet, it is not clear, whether these ingredients are the key for in-context learning to take shape in vision models. To stress-test such large models, we challenge them with an extreme counterexample: we train a tiny visual in-context model with merely $1$ million parameters and a modest amount of $70,000$ images. We compare the results of this severely capacity capped tiny model to $7,000\times$ larger VICL models in different adaptive settings, (1) on image data with small distribution shifts, (2) on unseen task encodings and (3) on a completely new task, i.e., the setting VICL envisions. With the chasm of training resources between the tiny- and large models, our experiments showcase a lack in how adaptive capabilities are measured, with respect to how tasks are encoded, which tasks were used in pre-training and the choice of metrics. These gaps in current VICL benchmarking underscore a need for innovation in evaluation of adaptive capabilities.

Monocular local navigation is attractive for lightweight robots, but existing vision-based policies often couple perception to a specific body, camera height, and footprint, making transfer from wheeled bases to legged platforms dependent on retraining or active depth hardware. This paper introduces AgniNav, a configuration-driven local navigation framework that standardizes cross-embodiment transfer at the collision-envelope level. Each robot is specified by a measurable four-parameter safety envelope: collision-relevant height, front length, rear length, and half width. The height parameter conditions an image-to-scan network to predict a one-dimensional, collision-relevant pseudo-laserscan from a monocular color image, while the remaining footprint parameters configure a dimension-aware local planner for collision checking. Training uses height-conditioned column-minimum scan labels generated from paired color-depth data, allowing the same image to supervise different safety envelopes without collecting robot-specific data. To the best of our knowledge, AgniNav is the first monocular local-navigation framework that jointly conditions perception and planning on a shared collision-envelope configuration for zero-retraining deployment across wheeled, quadruped, and humanoid platforms. Real-robot experiments on a Turtlebot2, Unitree Go2, and Accelerated Evolution K1 achieve 39/40, 18/20, and 18/20 successes with 0/40, 1/20, and 2/20 collisions, respectively, while running at 30 Hz on Jetson Orin.

Robotic manipulation has been widely applied in industrial scenarios. Compared with single-arm manipulation, bimanual manipulation is equipped with multiple cameras to capture information from different viewpoints. However, existing multi-view policies encode each view independently or fuse view features shallowly, resulting in limited sharing semantic perception and unreliable spatial awareness. In this paper, we propose \textbf{MV-Actor}, a multi-view perception framework that builds a unified semantic-spatial representation for bimanual manipulation. First, MV-Actor performs Multi-view Semantic Interaction to share semantic perception across views. Then it uses Semantic-Spatial Token Interaction to ground visual semantics with feed-forward reconstruction model features and acquire reliable spatial awareness. Finally, a Guided Metric Depth Repair module refines degraded sensor depth to provide more reliable metric anchors under consumer-grade depth noise. In simulation experiments conducted on the PerAct2 bimanual benchmark, MV-Actor achieves a state-of-the-art average success rate of 87.8\%. In real-world evaluations with more frequent viewpoint changes and unstable consumer-grade depth, MV-Actor outperforms both RGB and RGB-D baselines, further demonstrating the benefit of sharing semantic perception and reliable spatial awareness for bimanual manipulation.

Thermo-optic phase shifters are key building blocks in Silicon and Silicon Nitride-based reconfigurable photonic integrated circuits. They enable manipulating the phase of an optical signal by means of electrically-driven heating of an optical waveguide. Conventional fabrication schemes typically require dedicated lithographic steps to separately define the resistive heaters, the current transmission lines, and the electrical contact pads. This increases the process complexity and slows the standard complementary metal-oxide-semiconductor (CMOS) fabrication flows. In this work, we present a single-step lithographic process for the realization of Titanium Nitride thermo-optic phase shifters and Aluminum interconnects integrated on a Silicon Nitride photonic platform. A detailed electro-optical characterization, performed on two platforms operating at 810 nm and 1550 nm, revealed $π$-shift powers of 92 $\pm$ 2 mW and 120 $\pm$ 10 mW, respectively. Alongside, modulation bandwidths of 8.5 $\pm$ 0.3 kHz and 3.83 $\pm$ 0.03 kHz were extracted from combined frequency- and time-domain analyses. Our results demonstrate that the proposed single-step lithographic metal definition process represents a robust, viable and cost-efficient route towards CMOS-compatible reconfigurable Silicon Nitride photonics.

This paper presents an overview of the inaugural PortraitCraft Challenge, held as one of the official competitions at CVPR 2026. The challenge focuses on portrait composition understanding and generation, aiming to advance AI research in portrait aesthetics analysis and controllable image synthesis. Unlike existing datasets and tasks that primarily focus on global aesthetic scoring, PortraitCraft introduces a unified evaluation framework comprising two complementary tracks. Track 1 requires models to perform structured portrait composition understanding, and Track 2 requires models to generate portrait images from structured composition descriptions under explicit compositional constraints. To support the challenge, we constructed and publicly released a large-scale portrait composition dataset consisting of approximately 50,000 curated real portrait images, providing multi-level supervision. This report describes the challenge setup, evaluation protocols, dataset composition, and final results, along with an analysis of the technical characteristics of the submitted solutions. The PortraitCraft Challenge provides a standardized and reproducible platform for research on portrait composition understanding and generation, and is expected to foster further progress in the fields of portrait aesthetics and controllable image generation.

Post-training quantization (PTQ) compresses large language models by mapping weights to low-bit representations. The scaling factor that defines the quantization grid is typically chosen using simple, data-free heuristics. In this work, we present PiSO (Piecewise Scale Optimization), an algorithm that leverages calibration data to compute the optimal channel-wise weight scales exactly and efficiently under round-to-nearest quantization. PiSO partitions the scale search space into finitely many intervals on which the objective admits a closed-form minimizer. We extend PiSO to group-wise quantization via principled heuristics and propose effective strategies for interleaving scale optimization with error correction. Experiments on Llama and Qwen models across multiple model sizes and target weight bit-widths demonstrate consistent improvements in perplexity and downstream zero-shot accuracy, both standalone and combined with error correction. In particular, we observe increased benefits as the target bit-width narrows and quantization becomes more challenging.

Class-Incremental Learning (CIL) aims to continuously learn new classes without forgetting previously acquired knowledge. While recent CIL advances have spurred significant interest across various modalities, the audio-visual setting remains underexplored. Furthermore, although foundational multimodal models like SAM-Audio encapsulate rich static priors, our empirical analysis reveals that these representations struggle in incremental settings. This work bridges this gap by integrating SAM-Audio's audio-visual priors into the CIL setting. Specifically, we leverage its dense audio and visual representations and employ a novel guided attention strategy where the audio features contextually guide the visual representations. To further mitigate catastrophic forgetting, we introduce dual-level distillation objectives at both the feature and logit levels. Extensive evaluations on audio-visual CIL benchmarks demonstrate that our approach consistently outperforms state-of-the-art methods.

Large language models (LLMs) rely on tool use to act as autonomous agents, yet often fail in multi-step execution due to insufficient tool-related knowledge and ineffective knowledge activation. Therefore, we present a systematic study on how knowledge influences tool-use performance, covering the stages of knowledge acquisition, activation, and internalization. In the knowledge acquisition stage, we acquire and evaluate various forms of experiential knowledge, and our analysis shows that simple instance-level knowledge can already provide strong and reliable gains, while abstract intent-level knowledge offers limited benefits. At inference time, to activate knowledge, we find that prompting LLM to expand the depth of reasoning yields diminishing returns, whereas expanding the width of reasoning by parallel sampling with aggregation more effectively activates latent experiential knowledge. At training time, for knowledge internalization, post-training with knowledge-augmented data further improves performance, with reinforcement learning outperforming supervised fine-tuning. Based on these insights, we propose the Knowledge-Augmented Tool Execution (KATE), a knowledge-augmented tool execution framework that integrates experiential knowledge with reasoning-width-expanded inference and knowledge-aware training. Experiments on BFCL-V3 and AppWorld demonstrate consistent and substantial improvements over strong baselines across model scales. Our Code is available at https://github.com/hypasd-art/KATE.

Long-horizon autoregressive forecasting of oscillatory physical signals, such as seismograms, gravitational-wave strain, and similar wavefields is limited by error accumulation: as a causal model is fed its own outputs over hundreds of steps, small per-step errors compound into phase drift that pointwise metrics fail to detect. We ask when such rollout stays stable, using synthetic three-component seismograms as a physically structured testbed and the \textsc{SeismoGPT} autoregressive forecaster as the model under study. Through controlled, intra-architecture ablations evaluated on free-running rollout with paired significance tests, we isolate the contribution of each design choice. Multi-token prediction is the dominant stabilizer, accounting for almost the entire improvement over a single-token baseline ($+0.040$ median NCC); a horizon-embedding hybrid prediction head and a cross-horizon STFT-magnitude coherence loss each add a small but consistent further gain. Performance depends sharply on a context-ratio threshold near one, roughly the full P-S interval of observed signal, below which rollout generalization collapses. The dominant residual failure is a polarity inversion that a magnitude-based spectral loss cannot, by construction, penalize, identifying phase-aware objectives as the natural next step. We frame this as a controlled study of rollout stability on oscillatory wavefields, not a benchmark of forecasting architectures.

Vision-Language-Action (VLA) models achieve strong performance on standard manipulation benchmarks, but most evaluations assume that task-relevant objects are fully visible. This assumption often fails in realistic settings, where occlusion makes manipulation partially observable. In this paper, we study \textit{scene-induced occlusion} as a fundamental challenge for VLA models and introduce \textbf{LIBERO-Occ}, an occlusion-oriented extension of LIBERO. Experiments show that state-of-the-art VLAs suffer substantial performance degradation under occlusion. To address this issue, we propose \textbf{Viewpoint Imagination (VIM)}, which generates a complementary view from an occluded primary observation and conditions action prediction on both observed and imagined evidence. VIM improves robustness across task suites, occlusion types, and severity levels without requiring additional cameras at deployment time, suggesting that viewpoint imagination is an promising mechanism for perception completion in partially observable manipulation. Our benchmark and corresponding code are available at: \href{https://github.com/litsh/Libero-Occ}{https://github.com/litsh/Libero-Occ}.

LLM deception is often evaluated through direct markers such as fabricated claims, explicit lies, or strategic concealment. However, many real-world misleading communications do not depend on false statements, rather, they arise from selective treatment of true material facts: omitting adverse evidence, softening unfavorable details, emphasizing favorable details, or replacing precise qualifications with vague language. Existing benchmarks largely miss this subtler and arguably more dangerous failure mode. We introduce JANUS, a benchmark for measuring goal-conditioned pragmatic distortion in fact-grounded LLM outputs. Each scenario in our benchmark provides a fixed pool of favorable and adverse facts and compares a neutral condition against a goal-directed condition, such as increasing adoption, enrollment, approval, or support, despite potential harm to directly affected individuals or groups. Because all outputs are constrained to use the same fact pool, JANUS isolates misleading net impressions from hallucination and fabrication. JANUS contains 160 scenarios across 8 domains, with each scenario paired with neutral and goal-conditioned prompts and annotated material facts. Extensive experiments across 12 LLMs reveal consistent goal-conditioned distortions, demonstrating that current models remain sensitive to incentive and framing objectives and lack robust safeguards against selectively misleading communication. We publicly release our corpus and code for future research.

We describe ConvMemory v2, an opt-in token-evidence reranker that sits after the lightweight ConvMemory v1 reranker and reorders only v1's protected top-10 candidate set. v2 is a fine-tuned ms-marco-MiniLM-L-6-v2 cross-encoder (22,713,601 parameters, measured from the released checkpoint) applied to the ten (query, memory) pairs that v1 has already selected; it does not change which ten memories are returned, so Recall@10 and Hit@10 are identical to v1 by construction, not by statistical coincidence. On the LoCoMo conversational memory benchmark (5 seeds, n = 4955 test rows), v2 raises FULL MRR from v1's 0.5824 to 0.6560 (paired bootstrap +0.0734, 95% CI [+0.0645, +0.0827]) and H@1 from 0.4440 to 0.5474. v2 closes most but not all of the gap to a much more expensive full-pool cross-encoder reference (mxbai-rerank-large-v1 over the top-500, MRR 0.6688): on FULL MRR v2 sits 0.013 below mxbai_top500, but on two raw-dense-hard slices (where v1's protected top-10 has higher recall than mxbai's own top-10) v2 exceeds mxbai_top500. A four-arm load-bearing ablation shows candidate-specific memory text is the mechanism: removing, shuffling, or replacing it collapses MRR below raw dense retrieval. v2 is best understood as a standard recall-preserving cascade pattern with LoCoMo-specific fine-tuning, an explicit anti-shortcut inference contract, and disciplined load-bearing analysis; its advantage over mxbai is slice-specific rather than a general dominance claim. This report extends the v1 technical report (arXiv:2605.28062).

Current identity-consistent video generation methods struggle to preserve appearance fidelity under large viewpoint changes. While introducing multi-view reference input offers a natural solution, progress remains constrained by the lack of effective frameworks for multi-view inputs and the scarcity of multi-view data. We address these challenges by proposing HarmoView, a robust framework for identity-consistent video generation that effectively integrates multi-view cues through three architectural refinements complemented by a staged training curriculum. Specifically, we first introduce Multi-level Feature Injection to anchor identity fidelity; by injecting raw ViT features from frontal references alongside text tokens via cross-attention, MFI provides persistent low-level appearance anchors that complement the high-level identity features within DiT blocks, leading to enhanced identity preservation. Then, we employ learnable proxy tokens to unify heterogeneous reference layouts across single-/multi-view settings while simultaneously resolving the reference-view mismatch problem. Jump-RoPE is further developed for identity-wise feature isolation to reduce identity crosstalk. To activate these structural capabilities while preserving the original generative priors, we propose the Progressive View Curriculum. This four-stage training strategy employs view dropout to facilitate a stable transition from vanilla T2V generation to high-fidelity, identity-persistent spatial reasoning. Furthermore, we construct a large-scale multi-view dataset to address the issue of data scarcity. Extensive evaluation on our multi-view benchmark, comprising 100 manually-curated cases spanning 52 unique identities, demonstrates that HarmoView significantly outperforms open-source baselines and matches leading closed-source engines, achieving state-of-the-art performance in identity-consistent video generation.

Vision-Language Models (VLMs) demonstrate strong performance on general multimodal reasoning benchmarks, yet their ability to perform engineering reasoning remains largely unexplored. Unlike general visual question answering, engineering problem solving requires interpreting technical diagrams, selecting governing physical principles, and maintaining physically consistent multi-step reasoning. These capabilities are increasingly important for AI systems used in engineering education, scientific assistance, and technical decision-making, where reasoning failures may produce physically invalid yet superficially plausible solutions. Existing benchmarks primarily evaluate final answers and provide limited assessment of intermediate reasoning processes. We introduce EngVQA, a multimodal benchmark for evaluating engineering reasoning across 5 engineering subjects containing 696 problems. We introduce an 8-stage automatic evaluation framework for assessing VLM-generated solutions. The framework independently evaluates each stage of the solution, enabling fine-grained analysis of reasoning failures. We benchmark multiple state-of-the-art open and closed source VLMs on our evaluation framework and demonstrate substantial limitations in current engineering reasoning capabilities. Human evaluation shows strong agreement with our automated framework, achieving a Pearson correlation of 0.975 and a mean absolute error of 0.67 on a 10-point grading scale. Our results highlight the importance of process-oriented evaluation for reliable assessment of multimodal engineering reasoning systems.

Learning-based visual navigation for legged robots typically relies on continuous goal updates from hierarchical state estimation to provide a persistent directional reference. This reliance incurs additional sensory and computational overhead and deviates from fully end-to-end mobile autonomy. Furthermore, under partial observability, policies are prone to learn myopic behaviors, easily becoming trapped in dead ends and complex structural layouts. To address these limitations, we investigate a goal-initialized navigation setting, where the target is provided only once at the beginning of an episode, requiring the robot to operate based on intrinsic spatial memory without subsequent goal updates from external modules. In this work, we propose GUIDE, a fully end-to-end reinforcement learning framework designed to cultivate internal directional awareness. Specifically, GUIDE incorporates a spatial anchor predictor that leverages multi-frequency proprioceptive history to extract egomotion representations, thereby maintaining a persistent long-horizon spatial context for navigation. Concurrently, it utilizes raw depth streams to perceive local environmental geometry. We evaluate the proposed framework across both simulation and real-world scenarios on a quadruped robot. Experiments show that GUIDE learns reliable egomotion and directional awareness, enabling a fully end-to-end deployed policy to safely navigate through dense clutter and structured mazes without subsequent goal guidance or prior maps.

Diffusion policies (DPs) have emerged as expressive policy representations for robot learning, often used with imitation learning methods such as behavioral cloning (BC). However, while their success has largely been confined to BC, direct reinforcement learning (RL) fine-tuning remains challenging because actions are generated through a multi-step denoising process. In this work, we propose MODIP, a framework for the offline-to-online fine-tuning of DPs. Rather than directly applying RL to the DPs, MODIP leverages a world model (WM) to guide policy adaptation and keeps the simplicity and stability of BC. We utilize model predictive control (MPC) to generate high-quality trajectories within the WM, and use them as supervised targets for fine-tuning the DP. To make MPC planning efficient, MODIP uses a terminal state value instead of a policy-dependent state-action value, reducing inference time. Additionally, MODIP trains critics with policy-independent TD targets, reducing training time. Experiments on D4RL (MuJoCo, Kitchen) and RoboMimic tasks show that MODIP improves diffusion policies beyond BC, and is competitive with or outperforms diffusion policy RL fine-tuning methods and strong model-based baselines such as TD-MPC2.

The Euler Characteristic Curve (ECC) records the Euler characteristic of a linearly embedded cell complex as a function of filtration height in a given direction, and the Euler Characteristic Transform (ECT) is the injective shape descriptor obtained by collecting ECCs over many directions. How the ECT is encoded for a neural network is itself an inductive bias, conventionally fixed by discretizing each ECC. We introduce a continuous encoding: for each direction and each vertex it records the net Euler-characteristic change attributed to that vertex, producing a per-direction token sequence that a small transformer maps to a feature vector. We separate the resulting pipeline into two stages on orthogonal axes: an ECC encoder that acts within each direction, mapping its curve to a fixed-length vector, and an ECT representation that acts across directions, aggregating the per-direction vectors into one. We study six ECT representation architectures spanning a range of inductive biases, from a structure-agnostic feedforward baseline to convolutional and complex-valued models that preserve equivariance under planar rotations. Across six classification benchmarks covering point clouds, graphs, cubical complexes, and meshes, the continuous encoding improves accuracy on five of six datasets, and control experiments attribute the gain to the tokenization itself rather than to the added transformer capacity. The representation architecture matters less than the encoding, and the payoff from its inductive biases depends on the encoding: a feedforward network performs best under continuous encoding but is less robust under discretization than convolutional architectures.

RS-MLLMs enable natural-language understanding and spatial reasoning over earth observation imagery. However, existing models support only a narrow range of sensor types and tasks, yielding a fragmented view of the earth and leaving cross-modal geoscientific knowledge largely unexploited. This work presents Earth-OneVision, a 2B RS-MLLM that unifies six sensor modalities (i.e., optical, SAR, infrared, multispectral, temporal, and video) and cross-sensor fusion across 9 task categories within a single autoregressive framework. Three dedicated mechanisms address three bottlenecks. Full-Granularity Vision-Language Alignment (FGVLA) aligns multi-level visual features with the multi-dimensional language space. Spatial-Linguistic Isomorphic Serialization (SLIS) unifies heterogeneous spatial outputs as autoregressive tokens. Progressive Cross-Modality Adaptation (PCMA) decomposes the compound domain gap into sequential stages, tackling the viewpoint and imaging physics gaps in turn. To support joint training, MMRS-OneVision is constructed with ~34M QA pairs spanning all six sensor modalities and cross-sensor fusion across 9 task categories, substantially exceeding existing RS multimodal instruction datasets. With only 2B parameters, Earth-OneVision achieves competitive or state-of-the-art results across extensive benchmarks, consistently matching or outperforming 4B-72B RS-MLLMs. It achieves 87.52% P@0.5 on the OPT-RSVG testset for optical visual grounding and 80.68% on the SAR VQA benchmark SARLANG-Bench, exceeding 7B models by over 7%. It further achieves 75.74% recall on the BigEarthNet-MS testset for multispectral classification, and 81.94% MCQ accuracy on EarthMind-Bench for cross-modality reasoning.

Real-world robotic manipulation tasks often involve forceful interactions with the environment, such as using tools of varying weights, transporting objects with different masses, and performing contact-rich tasks like table wiping. Previous learning-based approaches typically employ imitation learning policies that output target end-effector poses tracked by low-level impedance controllers. In these systems, forceful interactions are either implicitly realized through steady-state tracking errors or explicitly commanded using wrist force/torque or tactile sensors. However, implicit approaches generalize poorly across object weights, while explicit approaches require specialized hardware and increase system complexity. In this work, we propose IMPACT, a framework that decouples these forceful tasks into task-planning and internal-model-based predictive control. Extensive simulation and real-world experiments demonstrate that the proposed framework achieves higher success rates and improved generalization to unseen object weights, as well as better safety and energy efficiency.

Users rely on execution traces to observe agent behavior, diagnose failures, and ensure accountability. These traces contain rich procedural detail, including tool invocations, intermediate decisions, and error-recovery logic. Yet this detail can expose private procedural skills, allowing downstream methods to recover key formulas, thresholds, and strategies without access to model weights or skill files. To quantify this risk and evaluate protection, we construct \textsc{CapTraceBench}, a benchmark of 75 specialized long-horizon tasks and 154 curated skills across seven domains. We also introduce \textsc{RedAct} https://github.com/XuShuwenn/RedAct, a protected trace release framework that localizes protected key information, rewrites traces while preserving verifier-critical evidence, and embeds behavioral watermarks for downstream provenance analysis. Across representative trace reuse methods, \textsc{RedAct} reduces normalized skill transfer (NST) from 44.7--67.1\% on raw traces to below the no-skill baseline, while preserving audit evidence. Its standalone behavioral watermarks reach 93.6--100.0\% true detection with a false alarm rate of at most 1.9\%. These results frame public agent traces as security interfaces and show that selective redaction can reduce procedural capability leakage without removing audit evidence.

Moonshine is an autonomous agent whose central objective is to generate mathematical conjectures. Its core capability is to extract structure from classical problems, distill new concepts, and formulate conjectures of mathematical significance. Rather than treating the solution of a single proposition as its endpoint, Moonshine builds an extensible theoretical framework through conjecture generation, bridge building, and obstacle identification. This article uses Moonshine's exploration of the Jacobian conjecture as an example. It shows how the central logic of whether local nondegeneracy can force global injectivity is transferred to one-hidden-layer affine-ridge sigmoid networks. This leads to the formulation of the \emph{Neural Jacobian Conjecture} (NJC): if such a network has strictly positive Jacobian determinant on the whole space, then it must be globally injective. By invoking GPT-5.5-pro and DeepSeek-V4-pro separately, Moonshine obtained independent complete proofs for the case \(N=n+1\). In addition, with the assistance of ChatGPT through interactive use of its web interface with GPT-5.5-pro, a geometric-topological proof was developed. These results provide preliminary evidence for the plausibility of the conjecture. The general higher-width case \(N\ge n+2\), however, remains unresolved and is left for further investigation. This work illustrates Moonshine's ability to autonomously generate meaningful mathematical problems and make rigorous progress on them.

Multimodal Large Language Models (MLLMs) excel at utilizing digital APIs and increasingly serve as the "brain" of embodied AI, instructing robots to interact with the physical world. In such embodied settings, a central capability is the use of physical tools, which underpins MLLMs' ability to assist humans in real-world tasks. Despite the importance, MLLMs' proficiency in physical tool use remains largely unexplored. To address this gap, we introduce PhysTool-Bench, the first physical tool-use benchmark designed to evaluate MLLMs' ability to comprehend real-world scenarios, identify physical tools, and plan their use. PhysTool-Bench comprises 2,510 queries over 2,678 real-world physical tools spanning diverse domains, including manufacturing, electrical work, agriculture, and healthcare. Concretely, models are evaluated along two primary dimensions: 1) recognizing all physical tools present in the scene, and 2) planning the tool selection and use sequence based on the instruction and visual context. Across 13 leading MLLMs, even the strongest model (Gemini-3.1-Pro) identifies only 58.7% of tools in a scene and completes merely 21.0% of queries end-to-end. Our analysis reveals a two-level deficit: MLLMs struggle to perceive tools in realistic scenes, and the much larger drop at the planning stage further indicates a lack of functional commonsense for mapping perceived tools onto task semantics, pinpointing a critical bottleneck for the development of practical embodied AI.

Large Language Models (LLMs) struggle to rigorously verify complex mathematical proofs. Standard global evaluation approaches suffer from "context poisoning," in which superficially plausible statements mask subtle logical flaws, leading to hallucination or over-skepticism. To address this, we shift from global evaluation to strict step-level verification: our framework maintains detailed context for each deduction step and strictly constrains the sources of applied theorems. We evaluate on a carefully curated adversarial diagnostic suite of research-level proofs drawn from the FirstProof challenge. A systematic ablation study demonstrates that these deductive constraints are indispensable, as unconstrained global prompting consistently fails to localize subtle logical errors. Beyond outperforming global evaluation, our approach fundamentally alters the failure taxonomy. Error analysis reveals that, rather than exhibiting severe logical hallucinations, remaining rejections are primarily instances of "pedantic hyper-rigor" stemming from unstated domain conventions, effectively exposing implicit ambiguities within the expert benchmark itself. Our findings suggest that prompting agents to organize their verification notes in a cautious, human-mathematician-like manner can substantially improve their ability to distinguish rigorous proofs from flawed ones, with the potential to strengthen agentic reasoning on frontier mathematical concepts that the base model does not already know well, and to lay a theoretical foundation for future automated proof-review systems. Code and prompts are available at GitHub.

Automatic Depression Detection (ADD) from clinical interviews is a pivotal task in computational mental health, yet it remains challenging due to two critical obstacles: 1) difficulty in modeling complex but sparsely distributed depression clues within lengthy, multi-topic clinical interviews, leading to superficial and unreliable reasoning; 2) scarcity of labeled data due to clinical privacy, together with high cost of training and fine-tuning, limiting the deployment of supervised ADD systems. To jointly address these challenges, we propose Dep-LLM, a training-free framework that mirrors the step-by-step reasoning of clinical psychiatrists and operates entirely on frozen off-the-shelf foundation LLMs. Dep-LLM comprises three stages. First, a Chain-of-Thought (CoT) Depression Multi-factor Analysis module structurally decomposes the long dialogue into five clinically aligned themes and produces evidence-grounded rationales, effectively handling long-context dependencies. Second, we introduce Confidence Analysis and Modulation module that quantifies the epistemic reliability from token-level entropy of each rationale and applies an intra-label and inter-theme modulation that amplifies trustworthy signals while suppressing uncertain ones without extra training. Third, a Collaborative Multi-factor Prediction module dynamically integrates multi-factor signals weighted by confidence into the final diagnosis. Extensive experiments on the DAIC-WOZ and E-DAIC datasets demonstrate the effectiveness and generalizability of Dep-LLM: it surpasses zero-shot baseline on nearly all 21 foundation LLMs across 9 metrics such as accuracy, macro F1 and weighted-average F1, and further outperforms state-of-the-art supervised domain-specific LLMs as well as the latest closed-source commercial LLMs, while requiring no extra training.

As agentic applications increasingly route user tasks through official and third-party LLM APIs, provenance becomes an operational question: which model generated a given black-box response? We study Dynamic Black-Box LLM Provenance: identifying the source LLM from generations elicited by query-varying, non-predefined prompts rather than a fixed input set or benchmark suite. This setting is difficult because prompt semantics dominate the text, while model-specific authorship traces are weak and inconsistent at the surface level. We introduce READER (Robust Evidence-based Authorship Decoding via Extracted Representations), a lightweight provenance framework that treats a frozen proxy LLM as a reader of hidden authorship evidence. READER maps black-box outputs into proxy activation space, temporally filters token states within each response, and performs Bayesian Evidence Accumulation by summing single-response log-posterior evidence across independently sampled prompts. This avoids fragile mean-pooling of prompt-specific representations while preserving the query-wise evidence needed for calibrated confidence. On Agent500, a 50-target dataset built from agent-style prompts, READER reaches $31.0$-$42.4\%$ top-1 accuracy from a single response and $70.0$-$84.0\%$ from 50 responses, substantially outperforming sentence-encoder fingerprints. Scaling across nine proxy readers further shows that stronger LLMs expose more linearly decodable authorship structure, suggesting that authorship perception is already present in frozen LLM representations and can be converted into reliable multi-query attribution.

Zero-shot learning (ZSL) for inertial measurement unit (IMU)-based human activity recognition (HAR) faces a central challenge: bridging the gap between sensor embeddings and semantic class representations. We systematically evaluate seven configurations combining three inference methods with two training pipelines on the PAMAP2 dataset, using 14 seen and 4 unseen activity classes with subjects 108 and 109 held out for testing. We find that the modality gap is a training-time phenomenon governed by the encoder objective. A temporal convolutional network (TCN) trained with cross-entropy over label-name Sentence- BERT prototypes yields sensor embeddings with a mean cosine similarity of 0.30 to the corresponding text prototypes, while replacing the label-name prototype targets with discriminative activity descriptions raises this to 0.69. This alignment improvement transfers consistently across all three inference methods. The strongest result combines contrastive training with inverted softmax correction, achieving 73.2% accuracy and 0.583 macro F1 on unseen classes, compared to 58.3% accuracy and 0.34 macro F1 for the label-name baseline. A secondary finding is that richer text descriptions reduce inter-prototype separability in Sentence-BERT space, because shared biomechanical vocabulary causes the language model to compress the prototype cloud. This effect does not negate the benefits of contrastive alignment provided prototype descriptions retain sufficient discriminative vocabulary. We also demonstrate that overall accuracy is a misleading primary metric when test-set class distributions are imbalanced, and recommend macro-averaged F1 as the standard reporting metric for ZSL-HAR benchmarks.

Foundation Models (FMs) have recently redefined the state-of-the-art in histopathology by providing robust representations for whole-slide image (WSI) analysis. However, selecting the optimal foundation model (FM) for a specific clinical cohort currently requires multiple preprocessing steps, followed by computationally expensive feature extraction and the training of a Multiple Instance Learning (MIL) aggregator for every model. In this work, we investigate whether efficient tile-level linear probing can serve as a reliable proxy for slide-level performance, reducing the need to run full slide-level pipelines for every candidate encoder. We benchmark 19 state-of-the-art FMs on 42 slide-level and 16 tile-level tasks, comparing tile probing metrics against slide-level outcomes using ABMIL and Mean Pooling aggregations. We observe a high correlation between tile and slide performance across varying task difficulties, indicating that encoder representation quality is the primary determinant of WSI success. Sensitivity analyses show that transferability is stable across models and is more influenced by cohort sizes and numbers of tiles per slide than by average task difficulty. We also measure the agreement in best performing models between tile and slide-level tasks, showing tile benchmarks reliably shortlist strong candidates. Overall, our study indicates that tile-level benchmarking provides an efficient and practical first step for narrowing down candidate models, while slide-level evaluation remains essential for final validation on clinical tasks.

The success of Large Language Models in mathematical reasoning relies heavily on the generation of diverse and valid solution paths during the rollout phase. However, current rollout techniques face a fundamental trade-off: token-level sampling often yields redundant trajectories that differ only in rephrasing, while embedding-level methods utilizing random noise frequently disrupt semantic consistency. To resolve this, we introduce N-GRPO, a novel exploration strategy integrated into the Group Relative Policy Optimization (GRPO) framework. Rather than relying on token-level sampling or native embedding-level noise, our approach leverages Semantic Neighbor Mixing. This mechanism dynamically constructs input representations by mixing the embeddings of an anchor token and its nearest semantic neighbors, thereby injecting diversity while strictly adhering to the local semantic manifold. Experimental evaluations on the DeepSeek-R1-Distill-Qwen models across different sizes show that N-GRPO not only achieves consistent improvements over strong baselines on math reasoning benchmarks but also exhibits robust generalization capabilities on out-of-distribution tasks.

Numerical solvers for partial differential equations (PDEs) are core computational tools in science and engineering. Building reliable PDE solvers requires not only executable code, but a numerical solver strategy, a set of decisions about discretization, stabilization, solver configuration, and resolution control, that matches the PDE structure. Recent LLM-based coding agents have begun to reduce the programming burden by generating and debugging solver implementations. However, they typically move directly from a PDE problem to solver code, leaving the solver strategy implicit in implementation details. Feedback from a failed solve is therefore routed back to code edits rather than to the underlying strategy, so numerical decisions remain hard to check before code is generated and hard to revise using numerical evidence when it fails. To address this limitation, we propose AutoPDE, a code agent that maintains the solver strategy as an explicitly represented object throughout the solving process: an independent, inspectable object that is built before any code is written and can be revised, using numerical evidence, whenever a solve fails. AutoPDE builds and maintains this object in three stages, all drawing from a library of reusable PDE-solving skills: PDE analysis identifies the equation type and algebraic structure; numerical method selection chooses a numerical method that matches the analysis result and commits to a discretization, stabilization, and linear solver accordingly; and adaptive tuning runs low-cost pilot solves to calibrate resolution and tolerances under the prescribed accuracy and runtime budget. We evaluate AutoPDE on the PDE Agent Bench, where experimental results show that AutoPDE achieves a pass rate of $54.5%$, improving over the strongest baseline by $14.2$ percentage points.

Large language model (LLM) agents are rapidly moving from conversational interfaces to software components that plan, invoke tools, maintain memory, and act on external environments. This transition changes the nature of security risk. In agentic settings, failures are no longer limited to unsafe text generation. Untrusted content may redirect control flow, misuse tool privileges, corrupt persistent state, leak sensitive information, or trigger harmful external actions. At the same time, research on LLM agent security is expanding quickly but remains fragmented across attack families, defense layers, application domains, and evaluation settings. This paper synthesizes 247 papers through a lifecycle-based, systems-oriented framework that models agent security around the interaction of information flow, delegated authority, and persistent state. We organize the literature around four questions: how LLM agent security should be modeled, which threat surfaces and attack families dominate, what defenses have been proposed and with what tradeoffs, and how security claims are evaluated. We find that prompt injection and tool-mediated control-flow hijacking still dominate the field, while persistent state corruption and multi-agent propagation are becoming central emerging concerns. We further find that current defenses provide useful building blocks but remain weakly compositional, and that existing benchmarks still underrepresent long-horizon, stateful, and deployment-sensitive risks. We argue that secure LLM agents require explicit trust boundaries, principled privilege control, provenance-aware state management, and evaluation practices aligned with realistic operational settings.

As AI systems built from multiple language-model agents become more common, they are increasingly used to make decisions together: discussing, negotiating, and acting on shared tasks. While individual agents may appear well-aligned when tested on their own, problems can arise from how they interact with one another. We introduce the Arbiter, an agent designed to monitor multi-agent conversations in real time and identify which participants may be behaving in misaligned ways. The Arbiter operates under a limited "inspection budget", meaning it must decide carefully how to use its resources. As it observes a conversation step by step, it can choose to wait, question a participant, examine internal information such as system prompts or reasoning traces, or log concerning behavior. At the end, it produces a report identifying the likely source of misalignment. We evaluate the Arbiter across five conversation conditions, ranging from risky financial advice model organisms to evaluation-aware and colluding agents, we test five tool configurations of increasing capability and two backbone models. We find that the Arbiter reliably detects misaligned agents well before the end of the conversation, with active inspection tools improving both detection accuracy and speed. Weight-induced misalignment proves hardest to detect, while instruction-induced misalignment is identified reliably even under passive observation. The logging tool exhibits a dual effect, improving recall at the cost of precision. These results suggest that continual, budget-aware monitoring can effectively catch misalignment, and that overseeing multi-agent systems may require treating the auditor as an active participant in the process. The code is available at https://github.com/aisilab/arbiter.

Robot Operating System 2 (ROS 2), the de facto standard middleware framework for robots, runs each robot as a graph of nodes communicating over the Data Distribution Service (DDS), a publish/subscribe substrate. Observing this inter-node communication in real time is essential to robot development, yet it has a price. A tool can receive data only by joining the DDS domain as a subscriber that discovery has matched to the publisher, so observing folds the tool into the system it measures and perturbs it. We define this protocol-inherent perturbation as the observer's probe effect. It inflates the discovery plane, adds deserialization cost on the observer, makes the loss it reports diverge from what the subscriber actually received, and near saturation displaces the subscriber's messages. The only escape, capturing all wire traffic passively, discards ROS 2 message semantics and scales with total traffic, not what is observed. We present ros2probe, a non-intrusive observation framework that removes the probe effect. It reconstructs the full ROS 2 communication state from the domain's discovery packets at no bandwidth cost, then drives an in-kernel filter restricted to the topics the user asks for, lifting only those packets at minimal cost and observing what the real subscriber receives. Its interfaces and recordings match the standard ROS 2 tools. Across three hardware platforms (laptop, Jetson, and Raspberry Pi), two DDS implementations, and seven robot-operation workloads, ros2probe holds the discovery graph within 0.5% of an unobserved system, whereas domain-joining tools inflate discovery up to 2.6$\times$ and drop 38.5% of the subscriber's messages at saturation while ros2probe drops none. It reports loss with a recall of 1.0, cuts observer CPU and memory by up to 7$\times$ and 28$\times$, and stays practical on the embedded robots where existing tools overload the system.

Learning from human video demonstrations remains challenging due to noisy hand-object interactions, unseen objects with partial observation, and cross-embodiment discrepancy. To address these challenges, we present \textit{HOWTransfer} (\emph{H}and-\emph{O}bject \emph{O}pen-\emph{W}orld Transfer), a hand-centric framework that distills human demonstrations into contact-aware, taxonomy-informed, and diverse robotic trajectories. Instead of relying on object-specific descriptions, vision-language queries, or explicit object-state tracking, \emph{HOWTransfer} recovers temporally consistent 3D hand motion and localizes temporal contact intervals by reasoning over observed hand-object interaction cues. The localized contact onsets are then used to retarget human grasp intent into multi-modal parallel-jaw grasp hypotheses, which are propagated along the recovered wrist trajectory to generate robot-executable motions. Finally, a trajectory editing stage refines contact alignment and produces diverse executable variants from a single demonstration. Experiments across diverse manipulation tasks show that \emph{HOWTransfer} enables accurate contact localization and high-quality robot motion retargeting with $86\%$ success, which is preferred over teleoperated trajectories in a blinded preference study.

External memory has become a core component of modern web agents, enabling long-horizon reasoning through the retrieval of past experiences. However, this paradigm introduces a critical vulnerability: malicious content injected into memory can be persistently recalled and repeatedly influence agent behavior. In this work, we identify and systematically study multimodal memory poisoning, an overlooked yet practical attack surface in web-agent systems. We propose MemVenom, a unified black-box attack framework that poisons graph-structured external memory with coordinated text-image evidence. Our method consists of a two-stage design: (1) a trigger-conditioned retrieval attack that ensures high-probability recall of malicious memory, and (2) a post-retrieval attack induction that leverages adversarial perturbations and stealthy OCR injection to override the original user objective. Unlike prior attacks that operate on prompts or text-only memory, our approach enables persistent, reusable, and goal-agnostic attacks without modifying model parameters or re-optimizing malicious tasks. Experiments across multiple web-agent frameworks and vision-language models demonstrate that MemVenom achieves strong end-to-end attack success with minimal impact on benign performance, reaching up to 99.15% on GPT-5-family web agents, while transferring effectively across architectures and model scales.

Large online courses generate thousands of student questions directed at conversational AI teaching assistants, yet these interaction logs remain largely untapped as diagnostic signals. We present a pipeline that maps student questions from a conversational AI teaching assistant to curriculum topics using a few-shot text classifier, grounded in a GPT-4-extracted prerequisite knowledge graph of course concepts. Evaluated on 1,340 question events from 164 students in a graduate-level AI course, our classifier achieves 80.0% accuracy across 43 labels (42 curriculum topics plus an "unknown" abstention class). Topic-level question volume correlates significantly with student self-reported difficulty from an independent mid-semester survey (rho = 0.491, p = 0.008, n = 28 topics), providing convergent evidence that the classified question stream reflects genuine topic difficulty. These results demonstrate that conversational AI interaction logs, mapped onto curriculum structure, carry actionable signals about topic-level knowledge gaps and provide instructors with a curriculum-grounded view of which topics warrant attention.

We study dense-to-sparse continual training as a way to construct channel-sparse large language models from dense checkpoints. Starting from a Qwen2.5-8B dense backbone, we continue training at 32K context and introduce a predictor-gated sparse SwiGLU FFN in the 32K stage. For each token and layer, we use a low-rank predictor to produce FFN-channel routing logits. We then apply a bank-wise top-k rule to retain 16 channels in every 64-channel bank, yielding 4x sparsity in the FFN intermediate activation. Unlike post-hoc sparse inference methods, the routing module is placed on the main language modeling path and optimized during continual training, enabling the dense model to be upcycled into a hardware-oriented sparse model. We report the architecture, training recipe, benchmark performance, and training lessons. We also identify a layer-local long-context failure mode on RULER-CWE and propose a single-layer repair algorithm that substantially improves the affected length range.

Pre-trained language models (PLMs) have achieved strong performance in keyphrase extraction (KPE), largely due to their ability to generate rich contextualized representations. However, long-document KPE remains challenging because salient keyphrase evidence may be scattered across distant document sections that cannot be jointly captured within the limited context window of most PLMs. Although long-context large language models (LLMs) can process broader textual contexts, their computational cost limits their practicality for efficient and high-throughput KPE. To overcome this limitation, we propose an attention expansion mechanism that augments PLM token representations with information from surrounding out-of-context chunks using pre-trained word embeddings. The proposed mechanism expands the effective contextual scope of PLM-based KPE models without requiring full-document attention or expensive LLM-based inference. We evaluate our approach across five PLM backbones, including general-purpose, scientific, task-specific, and long-context encoders, using two training regimes and five benchmark corpora from scientific and news domains. Experimental results demonstrate that attention expansion consistently enhances KPE performance across all evaluation settings, outperforming state-of-the-art models and yielding notable improvements in F1 score. The improvements extend to domain-specific, task-specialized, and native long-context models, showing that the proposed mechanism provides complementary information rather than merely compensating for limited input length. These results establish attention expansion as an efficient and effective strategy for long-document KPE.

The use of GRPO-style algorithms has become the standard strategy for training LLM search agents under outcome-only rewards. With these algorithms, a query contributes to parameter updates only when its rollout group mixes successes and failures; all-correct (too-easy) and all-incorrect (too-hard) groups are zero-variance and waste rollout cost. Existing approaches treat zero-variance as a static property and either discard or pre-filter such groups. We hypothesize and empirically validate that queries flip between zero-variance and signal-bearing states as the policy evolves during training. Building on this intuition, we propose query recycling, which returns zero-variance groups to a mutable pool for future resampling, so that the effective training distribution co-evolves with the policy. With the proposed technique, a 1.7B parameter model trained on synthetic data can reach 66.0 average Pass@1 accross seven multi-hop QA benchmarks, matching or surpassing systems with up to 7B parameters trained on benchmark-derived supervision. Analysis of recycling patterns shows that recycled queries supply roughly three quarters of the effective batch by the end of training, with contributions split between recovery from policy improvement and policy drift.

Resource constraints increasingly determine what can be trained, fine-tuned, and deployed in large language models (LLMs), yet efficiency is often studied through isolated techniques rather than as an interacting system of limits. This survey adopts a constraint-centric perspective and organizes recent progress around three coupled bottlenecks: data efficiency (what to train on), memory efficiency (how to fit training), and compute budget awareness (when and where to spend FLOPs). On the data axis, we review selection and pruning methods that maximize learning per token, ranging from scalable proxy signals based on learning dynamics to gradient- and influence-based scoring, as well as difficulty-aware and curriculum-style strategies. We highlight emerging evidence that different notions of good data dominate in different regimes, implying that optimal subsets depend on the task objective and resource budget rather than being universal. On the systems side, we show that GPU memory, not raw compute, is often the dominant bottleneck in fine-tuning, and that effective scaling requires jointly reducing weight storage, optimizer states, and activation memory rather than optimizing any single component in isolation. Beyond memory, we frame training and inference as compute-governed processes in which optimization, data selection, and decoding must explicitly account for finite FLOP budgets. We review evidence for compute-optimal allocation and stopping rules, where computation should be halted or reallocated once marginal performance gains fall below a budget-dependent threshold. Together, these results unify compute-aware data selection, scaling laws, and adaptive inference under a common principle of resource-conditioned decision-making.

In this paper, we use machine learning to discover a new seeding strategy for integration-by-parts reduction of Feynman integrals, which is a frequent bottleneck in state-of-the-art calculations in theoretical particle and gravitational-wave physics. Our strategy allows us to reduce multi-loop integrals with large numerator powers via essentially the standard Laporta algorithm but with a sparse selection of seed integrals that grows only linearly with the numerator power, whereas existing strategies lead to growth with a polynomial power that increases with the complexity of the integral being reduced. The seeds are restricted to a thin tube-like region that connects the target integral to the master integrals along a zigzag path. We demonstrate the power of our approach by reducing non-planar 2-loop 5-point integrals of rank 20 with numerical kinematics over a finite field, which is prohibitively difficult for the Laporta algorithm with conventional seeding. Going beyond individual integrals, we further demonstrate the reduction of a complete set of top-level rank-10 integrals by dividing the target integrals into several chunks, each of which can be solved by our sparse seeding strategy with considerably less time and a significantly lower memory footprint than other state-of-the-art strategies, making the approach well-suited for phenomenological applications. We provide a proof-of-principle implementation on GitHub at https://github.com/andreslunagodoy/tube_seeding.

Large Language Models (LLMs) are increasingly expected to interact with users over long time horizons. However, due to their finite context window, LLMs cannot retain all past interactions, making long-term memory management essential for storing, updating, and retrieving historical information beyond the context limit. Although recent memory systems attempt to address this issue by storing historical information externally, existing approaches suffer from three key limitations: flat text-based memory organizations fail to capture explicit relations among memories, structured memory systems often destructively overwrite evolving facts, and current retrieval mechanisms remain query-agnostic and passive when evidence is incomplete. REAL constructs long-term conversational memory as a temporal and confidence-aware directed property graph, where each atomic fact is represented with entities, relations, valid-time intervals, confidence scores, and exploration intent labels. During memory construction, REAL adopts a non-destructive temporal update strategy that preserves parallel fact versions and their validity intervals, enabling faithful tracking of fact evolution. During retrieval, REAL anchors query-relevant root entities, decouples their exploration intents, and performs semantic evaluator-guided hybrid beam search to extract compact memory subgraphs. It further incorporates counterfactual inference to repair unreliable retrieval states and recover missing memory evidence through implicit logical relations. Comprehensive experiments demonstrate that REAL substantially improves long-term memory performance over flat-text, graph-based, and existing memory baselines, achieving an average improvement of 22.72\%.

Modern language agents which perform multi-step reasoning have shown strong performance in knowledge-intensive question answering. However, existing approaches typically couple evidence acquisition and answer generation within a single policy. This forces a single model to play multiple potentially conflicting roles, inducing a combinatorial explosion in the policy space and hindering efficient exploration. It also introduces a credit assignment problem during training: a search action that retrieves sufficient evidence may still be penalized when generation fails, and vice versa. We propose DAC (Divide and Cooperate), a role-decomposed multi-agent training framework that divides agentic search into two cooperative subtasks, each handled by a dedicated agent trained with role-specific learning signals. The generator serves a dual role as both an answer producer and an evidence sufficiency verifier, abstaining when retrieved evidence is insufficient. This abstention signal is incorporated into the search agent's reward, providing structured cross-agent learning signals that improve credit assignment. Conversely, the searcher exposes the generator to diverse and challenging evidence environments by hard-positive evidence augmentation, improving its robustness. Experiments on general and multi-hop QA benchmarks show that DAC, implemented via parameter-efficient LoRA modules over a shared backbone, achieves strong performance against prior baselines that rely on full fine-tuning of monolithic models.

Dexterous hands are essential for fine-grained manipulation, but their hardware designs vary substantially across embodiments. Differences in kinematics, joint definitions, and degrees of freedom make it difficult to define a shared state representation compared with parallel grippers. As a result, dexterous-hand data remains fragmented and difficult to use for joint training. In this work, we propose the Unified Dexterous Hand Model (UDHM), which maps human and robot hand states into a shared 22-DoF semantic interface. Based on UDHM, we introduce UniDexTok, a retargeting-free state tokenizer that learns embodiment-conditioned discrete tokens from standardized real joint states. UniDexTok provides a unified representation for heterogeneous dexterous hands without relying on retargeting or simulation data. Compared with the recent baseline UniHM, UniDexTok reduces MPJAE from 15.63 degrees to 0.16 degrees and MPJPE from 18.51 mm to 0.18 mm, corresponding to error reductions of 98.98% and 99.03%, respectively. These results improve reconstruction from centimeter-scale to sub-millimeter accuracy. Experiments further show that data from other embodiments improves target-embodiment reconstruction accuracy, demonstrating the benefit of cross-embodiment tokenization. UniDexTok also shows strong zero-shot and few-shot reconstruction ability when new dexterous hands are introduced.

Transformer-based models have emerged as leading paradigms in time-series forecasting in recent years, employing self-attention mechanisms to capture long-range dependencies. Despite their success, these single-stage forecasting architectures exhibit persistent systematic residual biases arising from structural discrepancies, unmodeled stochastic components, or inadequate multi-scale temporal representations. This limitation persists when residuals are treated as irreducible noise, precluding adaptive correction of structured error patterns. To address this limitation, we introduce a two-stage, model-agnostic framework that explicitly decouples forecasting and residual learning into distinct stages of representation learning. A base transformer first generates the initial predictions. Subsequently, a dedicated meta-corrector dynamically models structured error patterns across multivariate channels, preserves cross-variable dependencies, and iteratively refines the residual bias of the base transformer. By formalizing this pipeline as a hypothesis space expansion, our framework addresses approximation limitations inherent in single-stage architectures, removes reliance on restrictive assumptions, and enables end-to-end learning of complex error dynamics. Evaluated on eight popular benchmark datasets using established protocols, our approach achieves state-of-the-art performance, with significant improvements in standard metrics (MSE, MAE). The results demonstrate the framework's ability to mitigate systematic biases and enhance robustness to complex temporal dynamics, advancing the practical applicability of transformer-based forecasting models.

Long-term LLM agents need persistent memory that can track changing facts and provide relevant evidence across sessions. Existing memory systems often store observations as isolated records, summaries, or indexed fragments, which makes evidence aggregation, fact revision, and memory maintenance difficult. We propose Infini Memory, a maintainable text-based persistent memory architecture that treats agent memory as topic-structured documents. Each topic document serves as a semantic unit for collecting related evidence, preserving metadata, and revising facts over time. New observations are first staged in a buffer and periodically consolidated into coherent textual contexts. At inference time, an agentic retrieval procedure lets the LLM read memory through iterative tool calls rather than a single retrieval step. On MemoryAgentBench, Infini Memory achieves 64.7% overall score. Ablations show that topic-structured maintenance and iterative evidence inspection improve complementary aspects of long-term memory use.

Although some very common test beds exist for assessing the performance of clustering methods, large scale benchmarking is typically limited to relatively simplistic simulation set-ups. Here we describe the production and curation of close to 3000 synthetic data sets, derived from more than 200 publicly available data sets; the majority of which arose from real-world applications. By fitting a flexible non-parametric distribution to each base data set we are able to retain much of the nuance in real-world data which is difficult to reproduce in standard simulations, while also producing data sets whose sizes are sometimes substantially greater than the data sets from which they are derived. The synthetic data sets, plus an accompanying R package, are available for download from https://github.com/DavidHofmeyr/ClusBench.

Autoregressive video generators synthesize long videos by generating successive temporal segments, but their historical KV cache grows with video length. Existing bounded-cache methods reduce this cost with local windows, sink tokens, or compressed memory states, yet they usually assign fixed roles to different parts of the history. We propose FadeMem, a distance-aware KV memory consolidation mechanism that organizes historical KV blocks into a temporal hierarchy under a fixed cache budget. This design is motivated by frequency-dependent temporal decay: fine details decorrelate quickly, while coarse scene structure and identity remain useful over longer horizons. During generation, new history is inserted as fine-grained entries, while older adjacent entries are progressively merged under a power-law temporal allocation schedule, yielding a dense-near, sparse-far memory within one cache. Without architectural changes, FadeMem preserves recent context for short-term dynamics and compact long-range anchors for identity and scene coherence. Experiments show improved subject consistency, background stability, and temporal coherence over existing bounded-cache strategies.

Multi-agent systems (MAS) can scale large language model reasoning at test time by decomposing complex problems into parallel subtasks. However, most existing MAS rely on centralized orchestration, where a main agent assigns work, collects outputs, and merges results. As the number of subtasks grows, this controller becomes a communication and integration bottleneck. We propose Decentralized Language Models (DeLM), a MAS framework that decentralizes coordination through parallel agents, a shared verified context, and a task queue. Agents asynchronously claim subtasks, read accumulated progress, perform local reasoning, and write back compact verified updates. The shared context acts as a common communication substrate, enabling agents to build on one another's verified progress without routing every update through a central controller. Empirically, DeLM improves both software-engineering test-time scaling and long-context reasoning. On SWE-bench Verified, DeLM achieves the best performance across Avg.@1, Pass@2, and Pass@4, with gains of up to 10.5 percentage points over the strongest baseline, while reducing cost per task by roughly 50%. On LongBench-v2 Multi-Doc QA, DeLM achieves the highest average accuracy across four frontier model families, improving over the strongest baseline by up to 5.7 percentage points. The code is available on our project website at https://yuzhenmao.github.io/DeLM/.

AI inference services -- API subscriptions, enterprise chat tools, and SaaS products with embedded AI features -- fall unambiguously within Scope 3 Category 1 under the Corporate Sustainability Reporting Directive (CSRD), which requires disclosure for fiscal years starting January 2024. Yet no standardised methodology exists for including them in corporate GHG inventories. Current practice either omits the category entirely or applies a generic economic input-output (EEIO) factor calibrated to the ICT sector as a whole, overestimating AI inference emissions by 10-40x relative to physically derived alternatives. We propose a four-tier framework that matches estimation precision to the data organisations can realistically obtain, progressing from direct token-based physical estimation -- using GPU energy benchmarks and regional grid carbon intensities -- down to a spend-based EEIO fallback for services where no usage data exists. Emission factors are derived from peer-reviewed GPU energy benchmarks (ML.ENERGY Leaderboard v3), confirmed grid carbon intensities (EPA eGRID 2023; Ember 2023), and published water use effectiveness data (Li et al., 2025). Applied to a 200-person European firm, the framework yields a total below 1 tCO2e, illustrating that the compliance challenge is methodological rather than magnitude-driven. We further document a water-carbon trade-off that current ESG tools do not surface: Sweden's hydro-dominated grid delivers the lowest carbon intensity in our dataset but the highest water footprint, with direct implications for data centre location strategy.

Multiple-choice (MCQA) benchmarks are the standard for evaluating pretrained large language models, but their reliance on log-likelihood scoring makes them unreliable. Specifically, standard scores are highly sensitive to the exact phrasing (surface form) of the answers, conflating a model's familiarity with a specific phrase with its actual capability. We demonstrate this flaw using a controlled testbed of 1B-8B models trained on the same knowledge. Despite having identical knowledge, standard metrics falsely report a performance gap of over 2 points. To solve this, we propose ParaEval, an evaluation framework that queries models using multiple paraphrases per answer option. By scoring each model based on its most favorable phrasing, ParaEval successfully reduces the false performance gap to below 1 point. We confirm that these evaluation artifacts, and the improvements from ParaEval, persist in frontier 70B and 120B open-source models. Ultimately, ParaEval provides a robust and efficient way to evaluate true underlying capability rather than surface-form familiarity.

We introduce Kwai Keye-VL-2.0-30B-A3B, an open-source Mixture-of-Experts (MoE) multimodal foundation model designed to advance long-video understanding and agentic intelligence. To address the challenges of ultra-long contexts, information redundancy, and prohibitive computational costs inherent in hour-level videos, Keye-VL-2.0 is the first to adapt DeepSeek Sparse Attention (DSA) to GQA-based multimodal architectures, enabling lossless 256K context processing while capturing critical frames and long-range temporal dependencies. This architecture is underpinned by a highly optimized training and inference infrastructure, including scalable video I/O, heterogeneous ViT-LM parallelism, and custom DSA kernels that significantly maximize throughput and minimize computational overhead. Furthermore, to overcome the algorithmic dilemma of catastrophic forgetting during multi-task alignment, we introduce Cross-Modal Multi-Teacher On-Policy Distillation (MOPD) paired with Context-RL and Video-RL. By distilling dense token-level teacher feedback from on-policy rollouts back into the MoE backbone, which activates only 3B parameters, Keye-VL-2.0 natively empowers advanced agent collaboration across Code, Tool, and Search scenarios with multimodal self-correction. Extensive evaluations across video understanding, temporal grounding, reasoning, STEM, and agent benchmarks demonstrate that Keye-VL-2.0-30B-A3B achieves state-of-the-art performance among models of similar scale, particularly excelling in fine-grained temporal localization on TimeLens and long-video comprehension on Video-MME-v2 and LongVideoBench. We release our model checkpoints to accelerate community progress toward scalable and robust multimodal agentic applications.

The scalability of Large Language Models (LLMs) to long contexts is fundamentally constrained by the quadratic complexity of standard attention, motivating the adoption of linear attention mechanisms with sub-quadratic cost. To improve representation capacity under long contexts, recent approaches organize memory in a multi-state manner. However, existing multi-state linear attention methods rely on fixed state merging policies that cannot adapt to dynamically varying token importance, irreversibly obscuring critical tokens and causing severe error accumulation over long sequences. To address this limitation, we propose DLA, a dynamic memory modeling framework for multi-state linear attention. DLA introduces (i) Information-Aware Dynamic State Merging, which adaptively determines state boundaries based on token-level information variation, preserving high-resolution representations around semantic transitions while aggressively summarizing stable regions, and (ii) Capacity-Bounded Memory Modeling, which maintains a fixed-size, chronologically ordered state cache by selectively merging adjacent low-information states to control memory growth with minimal information loss. We pre-train DLA on two different linear attention models and evaluate on 16 datasets across three categories. Experimental results demonstrate the superiority of DLA over state-of-the-art.

Reconstructing dynamic and interactive 3D scenes from real-world observations remains a fundamental challenge in computer vision and robotics. While recent advances in 3D Gaussian Splatting have enabled high-fidelity static reconstruction, extending it to interactive environments with articulated robots and manipulable objects remains difficult due to complex contact interactions and abrupt pose changes. To address these challenges, we introduce ManiSplat, a unified framework that reconstructs controllable and decoupled Gaussian digital twins directly from monocular ego-view robotic videos. Our method introduces a Graph-Structured Disentangled Representation that separates the robot, objects, and background into independently optimizable Gaussian subfields organized within a scene graph. To ensure stability, we propose a Task-Oriented Spatio-Temporal Alignment module that leverages the inherent logic of manipulation tasks-alternating between Motion and Skill phases-to construct accurate pseudo-ground-truth trajectories. Finally, a joint photometric-geometric optimization ensures the reconstructed scenes are temporally coherent, physically consistent, and simulation-ready. Extensive experiments demonstrate that our approach reconstructs interaction-driven dynamic scenes with high fidelity and controllability, effectively supporting downstream robotic tasks and policy learning.

Machine learning weather prediction (MLWP) models have achieved impressive forecasting performance at a small fraction of the computational costs required for traditional physics-based methods. However, they are primarily (1) data-driven and (2) evaluated using pixel-wide error metrics (e.g., RMSE), so there are no guarantees that their forecasts are consistent with known physical laws. We introduce PhysMetrics.Weather, an evaluation framework that assesses the physical realism of MLWP models across three types of metrics: conservation, spectral, and dynamical. By quantifying physical realism, this tool guides the development of physics-informed architectures and helps evaluate whether MLWP models are reliable for operational use. Our framework is available on Github at https://github.com/Emmakast/PhysMetrics.Weather.

Lipschitz-style individual fairness formalizes the idea that semantically similar examples should receive similar predictions, but its evaluation in multi-task learning (MTL) can be confounded by method-induced representation scales. This paper identifies threshold confounding: when the auditing tolerance is derived from each model's own representation distances, different algorithms are compared under different semantic thresholds. A threshold-drift analysis further shows how Bias rankings can change and identifies sufficient conditions for ranking preservation. We propose \textbf{ReLiF}, a reliability-aware framework that separates evaluation-time fixed-$δ$ auditing from training-time controlled regularization. ReLiF uses a shared reference tolerance for comparable auditing and a violation-rate feedback controller to keep the Lipschitz surrogate active without letting it dominate stochastic training. This work also develops supporting analysis for threshold drift, reference-tolerance selection, and the relationship between the huberized training surrogate and its unsmoothed positive-margin counterpart. Experiments on clinical time-series benchmarks and NYUv2 (NYU Depth V2) dense prediction show that fixed-$δ$ auditing exposes utility--fairness trade-offs that method-dependent thresholds can obscure. On NYUv2 with a ResNet50 backbone, ReLiF achieves competitive utility while substantially reducing aligned bias under shared fixed thresholds. On clinical benchmarks, ReLiF yields controlled fairness-regularized trade-offs, while fixed-$δ$ auditing reveals that task-balancing baselines can sometimes achieve lower bias and that genuine utility--fairness trade-offs persist. These results support fixed-$δ$ auditing as a semantically consistent protocol for evaluating Lipschitz fairness in MTL.

Image generation models now produce high-quality static images, yet their ability to represent how a visual world changes over time remains poorly understood. Practical workflows such as storyboarding, step-by-step illustration, reference-guided editing, and video previsualization require models to preserve identities, objects, spatial relations, and causal order across multiple visual states. Existing evaluations largely measure single-image correctness, compositional alignment, or video quality, leaving open whether an image model can coherently imagine a temporally ordered process. We introduce ImageTime, a diagnostic benchmark that uses spatiotemporal consistency as a behavioral probe of visual world modeling in image generation. Given an action instruction, and optionally a reference image specifying the initial state, a model must generate one image containing four ordered key states: initial state, action onset, transition state, and final state. This four-keyframe protocol is more temporally demanding than single-image generation while avoiding the confounds of dense video dynamics. ImageTime organizes tasks with a progressive capability hierarchy and decomposes each scenario into stage-wise state predicates, cross-frame temporal constraints, and forbidden causal violations. GPT-5.5 scores all generated images under a structured VLM-as-judge protocol, producing interpretable capability scores, diagnostic subscores, and failure labels. Through multi-family benchmarking, ImageTime reveals where current image generation systems succeed, fail, and drift when asked to maintain coherent visual world states over time.

Low-Rank Adaptation (LoRA) merging can efficiently combine diverse generative capabilities from multiple trained LoRAs for a diffusion model. However, existing LoRA merging techniques often suffer from severe parameter interference, causing destructive collisions in the shared parameter space. To address this, we propose Subspace Signal Routing (SSR), which resolves interference by routing internal signals instead of performing parameter-space merge. Specifically, SSR first constructs a unified subspace by concatenating candidate LoRAs along the rank dimension. Next, SSR employs an inverse correlation matrix to decorrelate mixed signals within this space. Finally, a directional guide matrix steers these purified signals into their respective task-specific subspaces. We provide a rigorous theoretical analysis proving that SSR aligns with the Ordinary Least Squares (OLS) solution, thereby ensuring mathematical optimality. We utilize the additivity of sufficient statistics to design a streaming algorithm. This enables on-the-fly updates that significantly reduce memory overhead and computation time. Extensive experiments validate that SSR significantly outperforms state-of-the-art methods while maintaining comparable efficiency. Code is available at https://github.com/nagara214/SSR-Merge.

Long-horizon language agents accumulate observations, reasoning traces, and retrieved facts that exceed their finite context windows, making memory retention a fundamental resource-allocation problem. Existing memory systems improve management through heuristic scoring, retrieval optimization, or learned compression, but largely treat retention as a local decision problem and do not explicitly model its long-term consequences under realistic observability constraints. To fill this gap, we formulate memory retention as a constrained stochastic optimization problem with explicit budget feasibility, evidence utility, and delayed costs including miss penalties, reacquisition delays, and stale-information risk. We then propose OSL-MR (Observability-Safe Learning for Memory Retention), a novel framework that enforces a strict separation between online-observable features and offline-available supervision (OAS). OSL-MR combines an evidence learner trained from realized evidence supervision with a Mixed-Score heuristic that serves both as a deployable online-safe baseline and as a structured inductive prior for learning. The resulting policy learns query-conditioned evidence value directly from interaction data while remaining deployable under the same observability constraints. Experiments on LOCOMO and LongMemEval show that OSL-MR consistently outperforms recency-based methods, Generative Agents-style scoring, and other heuristic baselines, particularly under tight memory budgets. The Mixed-Score prior further improves precision while preserving recall, and sensitivity analysis demonstrates robustness across a wide range of cost configurations.

Robotic foundation models pre-trained on human demonstration videos have shown promise, but a significant embodiment gap remains when the resulting policies are deployed on real robots. A common remedy is to fine-tune these models on robot-specific demonstrations. However, robot data collection can be prohibitively expensive and time-consuming, which is particularly acute in dexterous manipulation, e.g., teleoperating a multi-fingered hand for even a single atomic task can take days. To address this, we introduce Dexterous Point Policy, a framework that learns dexterous manipulation policies directly from human videos and requires no robot demonstrations. Our core insight is that a unified 3D keypoint representation can bridge human and robot embodiments when used for both observations and actions. Specifically, we extract 3D keypoints of task-relevant objects and human hands from raw videos, and train an autoregressive transformer over these keypoints. We observe that at the keypoint level, specifically the wrist and fingertips, human and robot behaviors closely align, enabling direct policy transfer. On a suite of real-robot tasks spanning pick-and-place and tool use, Dexterous Point Policy attains 75.0% success, whereas a state-of-the-art VLA baseline reaches only 1.0%. Furthermore, our method generalizes strongly to unseen scenarios, including multi-object environments and novel object categories.

Traditional heuristic solvers for the 2D irregular nesting problem share a fundamental limitation: they are blind to polygon geometry, relying on guided brute-force to navigate the continuous placement space with minimal geometrical guidance. In this paper, we argue that Reinforcement Learning is uniquely positioned to overcome this bottleneck. By pairing an optimization policy with a geometry-aware neural encoder, an agent can automatically discover rich geometric priors directly from data, utilizing these learned intuitions to strategically guide exploration. To realize this, we introduce the Polygons Transformer (PoT), a novel architecture that encodes 2D continuous vector geometries while allowing cross-polygons attention. We couple this novel architecture with a Combinatorial Optimization Reinforcement Learning (CORL) training framework to find optimal solutions. To support this paradigm, we release an open-source training dataset derived from complex geographic contours alongside a dedicated evaluation benchmark. Our empirical validation demonstrates that our trained agent achieves area utilization performance highly competitive with Sparrow, the state-of-the-art heuristic solver, proving that reinforcement learning can successfully discover and exploit geometric awareness for precise spatial tasks.

Parameter-Efficient Fine-Tuning (PEFT) has become essential for adapting foundation models to downstream NLP tasks. However, current PEFT methods often struggle with robustness to noise and performance degradation on limited training data. We propose SDBN (Small Data Big Noise), a unified framework that brings adversarial training to PEFT - a combination that remains less studied in the PEFT setting despite its complementary strengths - to enhance model robustness and generalization, outperforming alternative approaches. We also introduce two variants of the method that use discrete uncertainty sets: SDBN-h, which enumerates character-level edits and selects worst-case variants using gradients, and SDBN-p, which uses LLM-generated variants for robust optimization in generative tasks. Experiments across multiple benchmarks reveal substantial improvements, particularly in low-resource settings and under both word-level and character-level corruptions. This framework addresses the less explored intersection of adversarial training and parameter-efficient adaptation, without introducing additional parameters or only modest computational overhead, making PEFT deployments more reliable in real-world scenarios where data scarcity and linguistic variability often coexist

Causal discovery aims to uncover causal structures from observational data, which is crucial for real-world decision-making. However, different causal discovery algorithms can produce divergent results that conflict with each other, complicating the identification of accurate causal graphs. Traditional approaches rely on numerical values and statistical assumptions, often ignoring rich domain-specific information, such as feature descriptions, which could also help structure learning. While recent works explore using Large Language Models (LLMs) to infer causal relations via direct queries, such methods can be unreliable due to a lack of alignment with the actual data. To address these limitations, we propose Causal Ensemble Agent (CEA), a novel framework that aggregates structural insights from statistical discovery experts across different graph levels via linear opinion pooling, and uses an LLM as a meta-referee to dynamically reweight experts when the aggregated confidence is close to the decision boundary, thereby composing an improved and more complete causal graph. Extensive experiments on both synthetic and real-world datasets demonstrate that CEA achieves the strongest overall performance across a wide range of causal discovery methods, highlighting the effectiveness of using LLMs for meta-analysis in causal discovery.

Generating high-quality meshes for arbitrary geometries remains a fundamental bottleneck in computational engineering, often demanding heuristic tuning and semi-manual workflows. In this paper, we introduce Dmsh, a first fully automated reinforcement learning pipeline that unifies geometric decomposition and quadrilateral mesh generation within a single learning-based framework. Dmsh decomposes the problem through three coordinated agents handling topology simplification, geometric regularization, and mesh generation. The meshing process is formulated as a Markov Decision Process and solved using a parametric Soft Actor-Critic architecture with decoupled critics, enabling efficient exploration of a hybrid discrete-continuous action space. A curriculum learning strategy ensures scalability from simple domains to highly complex geometries, suppressing seed variance. By design, the recursive decomposition enables parallel meshing of subregions, yielding globally conforming all-quadrilateral meshes without post hoc correction. Across a wide range of benchmarks, Dmsh consistently outperforms existing methods in automation, robustness, and mesh quality, establishing a new paradigm for learning-based mesh generation.

3D vision-language segmentation aims to segment target objects in 3D scenarios according to the linguistic instructions and visual observations. Prior art heavily relies on the coarse superpoint representation to reduce the computation complexity, which suffers from poor segmentation quality and messy object boundaries. In this paper, we propose the SEGment-And-select (SEGA3D) paradigm for 3D visionlanguage segmentation that directly operates on the fine-grained visual information and is free from the superpoint dependency. Specifically, we first leverage a mask candidate generator to provide fine-grained categorical mask candidates, substantially improving the quality of candidate masks over the superpoint counterparts. Then, a Large Language Model (LLM) is utilized to generate the semantic and spatial information based on the linguistic description and visual features. The LLM output and visual features are fed to the Semantic-Spatial Selector (SSS) to produce the top-ranking mask candidates. Eventually, the Loopback Verification Module (LVM) is designed to yield the segmentation mask from the selected candidate masks. Our SEGA3D attains competitive performance on ScanRefer, ScanNet and Matterport3D benchmarks. Notably, our SEGA3D surpasses the top-performing counterpart by 8.3 mIoU and 5.3 mIoU on ScanNet and Matterport3D, respectively. Codes will be available upon publication.

Time series forecasting often suffers from over-smoothing, especially when future dynamics are multi-modal. Forecasts may follow the coarse trend of the observed future, but fail to preserve sharp changes, oscillations, turning points, and regime transitions that define plausible dynamic evolution. In this work, we revisit over-smoothing from the perspective of latent dynamical mode compression: under partial observation and single-realization supervision, multiple plausible future modes can be weakened, merged, or averaged during forecasting. Based on this view, we propose Dirichlet-Guided Group Forecasting (DGF), a mode-preserving forecasting framework that explicitly models multiple mode-conditioned predictive distributions and uncertainty over their selection probabilities. DGF uses a Dirichlet-guided hierarchical sampling mechanism and reward-based optimization to encourage forecasts that are accurate, dynamically consistent, and mode-distinct. Extensive experiments on real-world forecasting benchmarks show that DGF reduces over-smoothing while improving forecasting accuracy, diversity, and dynamical consistency.

Zero-shot vision-and-language navigation in continuous environments (VLN-CE) has recently become feasible with large vision-language models (VLMs). However, existing methods typically rely on learned waypoint predictors to propose navigable actions. This severely limits the model's action space and fails to leverage depth inputs effectively. Moreover, memory is commonly handled by accumulating long textual or visual histories with substantial irrelevant context, or by retrieving cross-episode experiences, which weakens the zero-shot setting. In this paper, we rethink zero-shot VLN-CE as an agentic interface between the VLM and the environment, and present AgenticNav, a lightweight navigation harness that exposes action, depth, and memory as callable tools. Instead of choosing from predicted waypoints, the action tool allows the VLM to directly select a target pixel in RGB observations, converting it into executable motion. Depth is exposed through an on-demand pixel-depth tool, enabling the VLM to request precise metric distances only where they matter. For memory, AgenticNav provides a compact map image summarizing the historical trajectory, paired with a recall tool that allows the VLM to selectively revisit past visual observations without overwhelming the prompt context. On the R2R-CE benchmark, AgenticNav establishes new state-of-the-art (SOTA) performance among zero-shot methods given the same VLM backbone. Real-world validation further highlights its zero-shot generalization compared to prior methods. Ablations show that our action tool design outperforms traditional waypoint predictors, and that depth tool and agentic memory further contribute to navigation performance.

Standard RLHF pipelines often reduce heterogeneous human judgments into a single scalar reward target. We argue that this reduction can mis-measure alignment in structurally plural societies, where disagreement may reflect culturally, historically, linguistically, regionally, or normatively grounded interpretations rather than annotation noise. We call this failure Preference-Validity Compression, the collapse of multiple plural-valid response options into a single optimization target. Using Malaysia as a diagnostic setting, we analyze RLHF-style feedback aggregation through preference events linking prompts, responses, and acceptability judgments across interpretive frames. Across 321 preference events from 20 participants and 107 trio-annotated prompts, 79% of prompts contain more than one majority-supported response that single-winner aggregation would discard, and apparent dominance gaps between top responses diminish when all majority-supported options are considered. Participants frequently select multiple acceptable responses, and discarded responses demonstrably reflect coherent local, practical, or cultural frames. These findings show that majority aggregation in this corpus measures argmax acceptability rather than plural alignment. We treat this as a measurement-validity issue and argue that future alignment methods should satisfy Validity-Preserving Consistency, remaining stable across plural-valid interpretive frames rather than collapsing them into a single reward target.

Vision-language-action (VLA) models have shown strong promise for robotic manipulation, but their reliability at test time remains limited by one-shot action prediction, where even small action errors can cause grasp failure, collision, or incorrect task progression. A natural alternative is to equip VLA systems with test-time verification, allowing multiple candidate actions to be proposed and evaluated before execution. However, reliable action verification is challenging because it requires not only distinguishing subtle geometric differences between candidate actions, but also assessing whether an action makes meaningful progress toward the task goal. We present VeriSpace, a 3D-aware action verifier for test-time action selection in VLA systems. VeriSpace evaluates candidate actions through two key components: Dual-Path 3D-Injected Scene Encoding, which constructs a scene representation that jointly preserves visual semantics and explicit 3D geometry, and Spatially-Grounded Action Reasoning, which evaluates each action by reasoning over task-relevant spatial relations, geometric validity, and expected goal progress. Together, these components enable more reliable discrimination between subtle yet outcome-critical action candidates while remaining fully compatible with existing VLA policies. Experiments on public benchmarks and real-world robotic manipulation tasks show that VeriSpace consistently improves decision reliability over both underlying VLA policies and prior verification-based methods, yielding substantial gains in both in-distribution and out-of-distribution settings.

Large Language Models (LLMs) are increasingly deployed in real-world applications that require access to up-to-date knowledge. However, retraining LLMs is computationally expensive. Therefore, knowledge editing techniques are crucial for maintaining current information and correcting erroneous assertions within pre-trained models. Current benchmarks for knowledge editing primarily focus on recalling edited facts, often neglecting their logical consequences. To address this limitation, we introduce a new benchmark designed to evaluate how knowledge editing methods handle the logical consequences of a single fact edit. Our benchmark extracts relevant logical rules from a knowledge graph for a given edit. Then, it generates multi-hop questions based on these rules to assess the impact on logical consequences. Our findings indicate that while existing knowledge editing approaches can accurately insert direct assertions into LLMs, they frequently fail to inject entailed knowledge. Specifically, experiments with popular methods like ROME and FT reveal a substantial performance gap, up to 24%, between evaluations on directly edited knowledge and on entailed knowledge. This highlights the critical need for semantics-aware evaluation frameworks in knowledge editing.

Skill documents, structured natural-language instructions that guide Large Language Model (LLM) agents, are critical to modern agent frameworks, yet LLMs struggle to write skills that actually work. On SkillsBench, human-authored skills improve pass rates by 16.2 percentage points, while LLM-authored skills provide no measurable gain. We introduce SkillAxe, a fully unsupervised framework that enables LLMs to iteratively diagnose and refine their own skills. SkillAxe decomposes skill quality into four interpretable dimensions (quality impact, trigger precision, instruction compliance with fault attribution, and solution-path coverage), producing structured improvement briefs that require no ground-truth labels, test suites, or environment rewards. On SkillsBench, SkillAxe improves pass rates by 28\% relative over unimproved LLM skills and closes 47--67\% of the gap to human-authored skills. We validate the approach as a continuous improvement engine in the wild on SpreadsheetBench, where a SkillAxe-built skill library learns from past agent trajectories and raises pass rate from 16.0\% to 52.0\% using only 22 skills.

Diffusion large language models (dLLMs) re-encode the entire prefix at every denoising step, causing recomputation that scales quadratically with context length and becomes prohibitive for long-context scenarios. We propose Prefilling-dLLM, a training-free prefill-decode disaggregation framework for dLLMs that partitions the prefix into N chunks, caches their KV representations once, and selects the top-K most relevant chunks with intra-chunk token sparsity for decoding, showing that sparse prefilling can outperform dense attention while reducing per-step complexity from quadratic in the full sequence length to quadratic only in the decode length. On LongBench and InfiniteBench, Prefilling-dLLM achieves state-of-the-art quality among dLLM acceleration methods, and an attention kernel that parallelizes decoding over the non-contiguously cached chunk KV yields 9.1--28.0x speedup at 8K--32K contexts. We further show that beginning-of-sequence tokens prepended to each chunk act as periodic attention anchors that eliminate the lost-in-the-middle phenomenon. Code is available at https://github.com/menik1126/Prefilling-dLLM.

Quantization-aware training (QAT) is essential for extremely low-bit large language models (LLMs). Current QAT methods are mainly based on scalar quantization (SQ), which enables efficient optimization but suffers from severe performance degradation at 2-bit precision. On the other hand, vector quantization (VQ) provides substantially higher representational capacity, but its discrete codebook lookup prevents end-to-end training. We propose LC-QAT, a 2-bit weight-only VQ-QAT framework that represents quantized weights via a learned affine mapping over discrete vectors, which yields a high-quality PTQ initialization and enables fully differentiable end-to-end optimization without explicit codebook lookup in the training forward pass. This strong post-training initialization makes LC-QAT highly data-efficient. Experiments across diverse LLMs demonstrate that LC-QAT consistently outperforms state-of-the-art QAT methods while using only 0.1%--10% of the training data. Our results establish LC-QAT as a practical and scalable solution for extreme low-bit model deployment.

Current reinforcement learning from human feedback (RLHF) methods primarily rely on scalar rewards from a trained reward model (RM). While effective, scalar rewards are often noisy and fail to capture fine-grained preference differences, whereas RM hidden states encode richer semantic and preference information. We introduce the representation-aware advantage estimation, which leverages RM hidden states and models them as auxiliary signals for better advantage estimation. Specifically, we propose the Graph-based Advantage Estimation (GraphAE), treat each sampled group as a graph, where nodes correspond to responses and edges capture their similarity in the RM hidden space. Then advantages are computed via graph propagation, enabling each sample to incorporate contextual information from its neighbors. GraphAE is lightweight and can be seamlessly integrated into existing group-based RL algorithms. We apply GraphAE to GRPO, GSPO and RLOO, and conduct extensive experiments on different models and benchmarks. Empirical results show consistent improvements across three benchmarks, with gains of up to + 6.3 on Arena-Hard-v0.1, + 8.27 on AlpacaEval 2.0, and + 0.22 on MT-Bench. These results demonstrate that leveraging RM representations leads to more sample efficient and robust RLHF.

Post-training quantization at the 2-bit level enables low-cost deployment and inference acceleration for large language models (LLMs). Scalar quantization (SQ) and vector quantization (VQ) are two primary quantization methods, however, the former suffers from significant performance degradation, and the latter incurs computational and storage overhead. We propose UniSVQ, a unified 2-bit quantization framework that bridges scalar and vector quantization by parameterizing codewords as an affine transform of integer lattices. This structure preserves compatibility with optimized integer kernels while retaining much of VQ's flexibility. We further introduce a data-driven block-wise fine-tuning strategy to directly minimize quantization reconstruction error. Extensive experiments across multiple LLM families and zero-shot benchmarks demonstrate that UniSVQ consistently outperforms state-of-the-art SQ methods and achieves performance comparable to advanced VQ methods, while providing higher inference throughput.

Deploying Vision-Language-Action (VLA) models in real robotic systems requires robustness not only to semantic and perceptual variations, but also to embodiment-side faults that change how actions are physically realized. Real robots can experience joint-level changes caused by actuator degradation, hardware faults, safety limits, collision damage, or wear-induced friction. These faults are critical because they alter the action-to-motion interface of a policy, disrupting the learned closed-loop relationship between commanded actions, realized motion, and subsequent observations. In this work, we study realistic joint-level physical faults and show that VLA models are vulnerable when predicted actions are executed through a perturbed robot body. Our analysis reveals joint-dependent effects, with heterogeneous degradation in task success across affected joints. We also show that performance drops cannot be attributed solely to physical infeasibility, since feasible faults such as increased joint friction can still substantially reduce success rates and induce closed-loop execution mismatch. Motivated by these findings, we propose Joint-level Physical-fault Aware Residual Calibrator (J-PARC), a lightweight residual calibration framework built on top of a frozen VLA policy. J-PARC infers a latent joint-fault regime from recent joint dynamics and conditions a shared residual calibrator on this regime, enabling adaptive action correction across faulty joints. Experiments show that J-PARC improves robustness under joint-level faults while preserving fault-free environment performance.

Safe social navigation requires robots to distinguish people from ordinary obstacles and to react before danger becomes imminent. We show that pretrained Vision-Language-Action (VLA) models already encode pedestrian-object distinctions and future collision signals in their internal representations, but behavior cloning fails to translate these signals into socially appropriate actions. To address this mismatch, we propose SALSA, a two-stage annotation-free post-training framework: (1) social behavioral alignment bridges intermediate-layer social features to the action head and trains on counterfactual human-object scene pairs to break visual saliency shortcuts; (2) temporal safety alignment provides automatically generated future-risk supervision to enable anticipatory collision avoidance. On SCAND and real-world deployment, SALSA reduces near-collisions by 86.4% and improves social counterfactual accuracy from 53% to 93%, demonstrating that safer social navigation can be achieved by teaching VLA policies to act on representations they already possess. These results show that pretrained VLA policies can be adapted for safer social navigation by better aligning their latent representations with action generation.

Most recent 3D reconstruction and editing systems operate on implicit and explicit representations such as NeRF, point clouds, or meshes. While these representations enable high-fidelity rendering, they are fundamentally low-level and hard to control programmatically. In contrast, we propose and systematically evaluate a new 3D reconstruction paradigm, 3D Code Synthesis (3D-CoS), where 3D assets are constructed as executable Blender code, a programmatic and interpretable medium. To assess how well current VLMs can use code to represent 3D objects, we evaluate representative open-source and closed-source VLMs in code-based reconstruction under a unified protocol. We further introduce a suite of structured code-synthesis workflows, including blueprint-based planning, Retrieval-Augmented Generation (RAG) over Blender API documentation, few-shot geometric demonstrations, and a component-level Agent workflow for part-wise code generation. To demonstrate the unique advantages of this representation, we further evaluate localized text-driven modifications and compare our code-based edits with a point-cloud-based 3D editing baseline. Our study shows that code as a 3D representation offers strong controllability and locality, yielding stronger edit fidelity and better preservation of unedited regions in our targeted editing evaluation. Our work also analyzes the potential of this paradigm, delineates the current capability frontier of VLMs for programmatic 3D modeling, and highlights code synthesis as a promising direction for editable 3D reconstruction.

Multi-agent debate frameworks have been shown to improve large language model performance in convergent tasks, but they are currently optimized in a way that heavily favors final output accuracy rather than stability of the process. During long-horizon exchanges reactive systems under sustained perturbations often experience logic degradation, argument repetition, and role drift. To structurally prevent the identity loss and maintain the process fidelity, we introduce Knowledge-Grounded Counterfactual Reasoning (KG-CFR), a dual-stage architecture that enforces a strict separation of concerns between a private, retrieval-augmented planning buffer, and a public execution layer. We assess this system in Dynamic Resource Allocation under Uncertainty (DRAU), a dedicated 1v1v1 environment, introducing diversity as distinct from standard debate settings. Over 270 completely factorial crisis simulation trajectories with stochastic environmental shocks, KG-CFR prevents judge-detected critical post-shock degradation (defined as a quality shift, $Δ\le -0.20$) in more than 95% of perturbed runs, increasing the overall argument quality from 0.694 to 0.822. Our primary contribution is the demonstration of architectural decoupling being an important factor of systemic resilience enhancement under sustained pressure without quality loss. Furthermore, we introduce custom vector metrics for discourse divergence and plan-execution alignment that provide strong, directionally consistent evidence of operational stability. Our ablation experiments suggest that the proper doctrinal grounding can be an equally important factor for argument quality, as the prospective planning. KG-CFR, according to our initial metric evaluations, reduces semantic looping, by preserving the agent's consistency with the original plan.

Recent large language models (LLMs) have shown rapid progress in reading-based question answering (QA), where evidence is explicitly provided or can be trivially retrieved. In contrast, real-world questions are often not paired with accurate evidence documents. The useful evidence resides in massive data lakes, making search a prerequisite for answering. However, there is a lack of comprehensive benchmarks that require both searching and reasoning over large data lakes. To this end, we introduce LakeQA, a comprehensive benchmark for search-centric question answering over data lakes that jointly emphasizes searching and reasoning capabilities. LakeQA is built on a heterogeneous collection of approximately 9.5 TB of text resources from Wikipedia and open-source government data, spanning structured and unstructured data. To ensure task quality, each sample is annotated by at least one Ph.D.-level expert. Each task requires long-horizon multi-hop reasoning with implicit intermediate steps: agents need to discover the correct documents and then compose evidence across sources to produce the answer. Experimental results on seven frontier LLMs demonstrate that LakeQA is challenging. For instance, GPT-5.2 achieves only an exact-match score of 18.37% on LakeQA. Overall, LakeQA provides a realistic testbed for developing LLM agents that can both find and analyze data in modern data lakes.

DiffC provides a principled way to reuse pre-trained diffusion models for lossy compression, but its encoding and decoding procedures remain slow because they require many discretized forward and reverse steps. We study whether few-step generative models -- Rectified Flow, Consistency Trajectory Models (CTM), and MeanFlow -- can be cast as codecs within the same reverse channel coding (RCC) framework. The main challenge is that RCC requires posterior and shared distribution parameters, whereas these models do not explicitly parameterize intermediate conditional distributions. For Rectified Flow and MeanFlow, we use the equivalence between velocity parameterization and diffusion-style denoising parameterization to derive the quantities required by RCC. For CTM, which is distilled from EDM, we adopt the EDM noise parameterization together with local Gaussian approximations of the sender and shared distributions at intermediate states. This yields a proof-of-concept probabilistic formulation that enables compression with pre-trained few-step generative models without retraining. On low-resolution benchmarks, the resulting codecs reduce encoding and decoding time and improve realism in the low-bit-rate regime.

Humanoid robots have achieved strong locomotion capabilities, but reliable navigation on versatile terrains remains challenging because obstacle avoidance must be coordinated with dynamically feasible motion. In this work, we present GuideWalk, a unified end-to-end framework that integrates traversability-aware navigation guidance with terrain-adaptive locomotion teacher for humanoid navigation. Specifically, we introduce a navigation module that provides explicit velocity guidance, decoupling obstacle avoidance from terrain conditions to enable robust planning across diverse environments. We propose a composite teacher distillation scheme, where goal-directed commands and dynamically consistent actions are aggregated and distilled into a single policy. To further improve robustness, the distilled policy is refined with reinforcement learning and an auxiliary behavior cloning objective, which promotes exploration while preserving desirable teacher behaviors. Experiments demonstrate that GuideWalk achieves stable and effective navigation while maintaining stable humanoid locomotion.

Semi-structured 2:4 sparsity is widely supported by modern accelerators, providing up to a 2x theoretical speedup. However, its strict 50% sparsity constraint often causes non-negligible accuracy degradation under post-training pruning. Meanwhile, existing relaxed sparsity formats either require specialized compiler support or introduce runtime overheads that limit end-to-end speedup. We propose Spense, a practical hybrid sparse-dense format that splits each weight matrix into a 2:4 sparse region and a dense region. This design relaxes the effective sparsity constraint while remaining compatible with existing high-performance sparse and dense GEMM libraries, avoiding both custom compiler support and input activation expansion. Building on this format, we introduce SpenseGPT, a one-shot post-training pruning method that produces sparse and dense regions. Notably, we show that selecting the right dense regions is important, and we devise two different strategies to choose them. Experiments on Qwen3-32B and Seed-OSS-36B demonstrate that our method achieves up to 1.2x end-to-end decoding speedup on B200 GPUs with FP8 precision, while preserving accuracy. To the best of our knowledge, this is the first one-shot pruning demonstration of real-world end-to-end LLM decoding speedup from semi-structured sparse tensor cores on recent GPUs such as B200s, while maintaining model quality.

Transformers achieve strong language modeling performance by providing direct token-to-token communication paths, but causal self-attention scales quadratically with context length. Recurrent and state-space models reduce this cost, yet compress history into sequentially updated fixed-size states. This paper studies a third primitive: a parallel content-addressed memory over causal successor records. The proposed Parallel Causal Associative Field (PCAF) writes local records from a context window into hash buckets, retrieves a bounded candidate set for the current query, forms a sparse cache distribution over successor tokens, and mixes that cache with a parametric local language model through a learned gate. The resulting model maintains sparse long-context access while avoiding a single fixed recurrent state bottleneck. We evaluate PCAF under full autoregressive pretraining on WikiText-103 and PG-19 using a distributed Google Cloud TPU v4-32 pod. At 303M parameters and context length T = 2048, PCAF-semantic reaches 36.31 perplexity on WikiText-103 and 52.45 perplexity on PG-19, compared with 47.49 and 53.84 for a matched dense Transformer. PCAF-semantic simultaneously processes 0.61-0.62M tokens/s across the TPU pod, versus 0.43M tokens/s for dense and local attention baselines. Supporting 41M-parameter multi-seed sweeps and single-GPU component ablations show that the associative cache, retrieval capacity, and learned gate materially affect the speed-quality trade-off.

Autonomous web navigation remains challenging for LLM agents, and the strongest generalist systems rely on proprietary reasoning models whose inference cost is prohibitive for the repetitive tasks where such agents would be most useful. We argue this gap stems not from insufficient model capability but from agent architectures that fail to replicate three human cognitive advantages: selective attention to relevant page regions, persistent memory of website structure, and procedural fluency with common interaction patterns. We introduce WebChallenger, a web agent framework that addresses each gap through architecture design rather than model scale, built around PageMem: a structured page representation deterministically constructed from the DOM that exposes each page as a hierarchy of semantic sections with short summaries. On this shared substrate we build three mechanisms that mirror the three cognitive advantages: a divide-and-conquer observation pipeline that lets the agent skim section summaries and extract details only from task-relevant regions; a lightweight exploration and memory system that traverses each website once to build a reusable map of pages and element behaviors; and compound action workflows that collapse common multi-step interactions into single agent actions, handling partial state changes automatically. Because all three operate over PageMem, the framework generalizes across websites without site-specific adapters. Using off-the-shelf open-weight models without fine-tuning, our system achieves 56.3% on WebArena, 48.7% on VisualWebArena, 51.0% on Online-Mind2Web, and 70.9% on WorkArena, approaching frontier proprietary systems at a fraction of the cost. Our code is released at https://github.com/jayoohwang1/webchallenger

Math reasoning benchmarks have proliferated, yet most lack a per-item difficulty signal grounded in actual human performance. We introduce KCSAT-ML, a decade (2014-2025) of Korean College Scholastic Ability Test (KCSAT; Suneung) mathematics: 664 problems with a 339-item core set carrying official per-item error rates from nationwide cohorts of hundreds of thousands of examinees. We pair the benchmark with Difficulty-aligned Reasoning Gain (DRG): a score-orthogonal metric that asks whether a model's mistakes concentrate on the items humans found hard, or on items humans found easy. Together they expose, across a wide range of VLMs (and LLMs via OCR), three patterns: (i) low-budget accuracy collapses on the high-human-error tail at every model size; (ii) test-time scaling (TTS) raises token use roughly linearly with cohort error rate, while accuracy gains follow a non-monotonic curve; (iii) within a single family, TTS flips between anti-scaling on the hardest items and overthinking on easier ones -- two faces of the same alignment failure. On DRG, models with near-identical accuracy can sit at near-opposite values: one model gets wrong what humans also find hard, while another solves the hardest items yet fails on items humans find easy -- a contrast that aggregate accuracy hides. Our code and dataset builder will be open-sourced at https://github.com/naver-ai/KCSAT-ML.

Scientific discovery is often a collective process: researchers share partial results, inspect failed attempts, and build on each other's ideas over long time horizons. Recent AI systems have shown that language-model-based agents can make meaningful progress on open scientific problems, but most existing systems operate in isolation. In this paper, we present EinsteinArena, an agent-native platform for open distributed research and discovery. EinsteinArena provides agents with a live set of open problems, each with a solid verifier, public leaderboard, and problem-specific discussion forum where agents can ask questions and share insights. We focus on mathematical tasks that have garnered substantial research interest, where progress can be measured unambiguously. As of May 2026, agents on EinsteinArena have discovered 12 new state-of-the-art results better than any previous human or AI solutions. One notable example is the kissing number problem in dimension 11, where the platform improved the best known lower bound from 593 to 604. This advance did not come from a single agent or isolated run. Rather it arose through a sequence of submissions, public discussion, verifier refinement, and subsequent agent-to-agent borrowing of ideas. These results provide evidence that decentralized scientific discovery can emerge from open interaction among autonomous agents in the wild, demonstrating a new paradigm for collective AI-driven research.

Spatial intelligence is a key frontier for multimodal large language models (MLLMs), enabling them to reason about the physical world from visual experience. Inspired by human spatial cognition, recent approaches construct grid-based cognitive maps from multi-frame visual inputs to maintain coherent spatial representations over time. However, limited context lengths still challenge spatial understanding, while existing methods, such as long-context modeling and external memory, often require architectural changes, memory modules, or finetuning, limiting their applicability to off-the-shelf pretrained MLLMs. This motivates a lightweight, model-agnostic method for preserving spatial information beyond the native context window. To this end, we propose a plug-and-play multi-agent framework that collaboratively constructs cognitive maps as structured spatial memory, enhancing the spatial understanding of arbitrary pretrained MLLMs without architectural modification or additional training. Our framework features local-global agent coordination, cognitive map construction with atomic commits, and cross-agent verification. Extensive experiments demonstrate that our method achieves superior performance on spatial understanding tasks while remaining fully training-free. Code will be released.

Vision-language models (VLMs) are increasingly deployed where answers must follow from what is in the image, yet they often answer from textual priors, the question's phrasing together with memorized world knowledge, rather than from the image itself, which inflates benchmark scores and yields confident but ungrounded answers. Existing benchmarks rarely isolate this behavior, since each image is usually paired with a single fixed question. To measure the reliance, we build a 540-image benchmark across six reasoning categories and generate four question variants over the same images, so that phrasing rather than image content is the controlled variable. The hardest variant is written directly from the image to minimize text leakage. We benchmark eleven VLMs spanning small open-weight models to large closed-source systems: every model degrades on the hardest variant, and open models fall furthest. Our central diagnostic is a no-image ablation, which collapses the open-weight models to their text-only floor (1 to 9 percent). Three further analyses, LLM-rated difficulty, low base-to-final textual similarity, and human re-annotation, corroborate genuine image-dependence. In-context exemplars that match how a variant was built recover the most accuracy, and GRPO post-training of a small VLM yields consistent gains across all four variants that transfer to a held-out out-of-distribution set. Textual-prior reliance is measurable and partly trainable away.

Real-robot evaluation is essential for understanding whether learned manipulation policies can operate reliably outside curated demonstrations. This need is particularly pressing for Universal Manipulation Interface (UMI)-style policies, whose performance depends on the coupling between wrist-view observations, action representation, data collection, and physical deployment. Existing real-world benchmarks have made important progress, but they are not designed around this UMI data-to-deployment setting. We present UMI-Bench 1.0, a local-first real-robot benchmark for standardized evaluation of UMI-style manipulation policies. To the best of our knowledge, this is the first benchmark dedicated to real-world evaluation of UMI-based manipulation models. UMI-Bench aligns data collection, scene reset, policy execution, result logging, and task-factor analysis within a unified protocol. By making the full evaluation process reproducible and auditable, UMI-Bench provides a practical testbed for measuring how UMI-trained policies generalize to real physical manipulation.

Muon collider research spans accelerator physics, detector instrumentation, and high-energy phenomenology, with relevant evidence scattered across a rapidly expanding and heterogeneous body of scientific literature. As high-energy physics (HEP) increasingly explores agent-assisted analysis workflows, efficiently locating, integrating, and verifying scientific evidence becomes an essential capability. While retrieval-augmented generation (RAG) offers a promising framework for scientific question answering, integrating agentic reasoning without compromising retrieval precision remains a key challenge. In this work, we present agentic hybrid RAG, an evidence-grounded RAG framework for muon collider research. The framework combines a hybrid retriever, integrating sparse lexical and dense semantic retrieval, with an agentic reasoning module for query decomposition, evidence expansion, and grounded answer generation. To enable systematic evaluation, we construct the first benchmark for retrieval-augmented scientific question answering in the muon collider domain, comprising a curated literature corpus together with dedicated retrieval and answer-generation benchmarks covering major detector and physics research topics. Extensive evaluation shows that hybrid retrieval provides the strongest retrieval backbone, while agentic reasoning is most effective for controlled evidence expansion and answer synthesis. Built on this principle, agentic hybrid RAG consistently outperforms representative retrieval and RAG baselines in retrieval effectiveness, answer quality, evidence coverage, and factual grounding. Together, the benchmark and framework provide a foundation for evidence-grounded scientific question answering and future HEP analysis agents operating over large-scale scientific literature.

Real-world crisis intervention is inherently conversational, yet existing research largely focuses on static texts.Real-world crisis intervention is inherently conversational, yet existing research largely focuses on static texts. When applied to multi-turn dialogues, current models exhibit significant performance degradation, struggling to track risk signals that emerge as context evolves. To address this gap, we introduce CRADLE-Dialogue, a clinician-annotated benchmark for turn-level crisis detection in conversational settings. The dataset features 600 dialogues with multi-label annotations across clinically grounded risks, including suicide ideation, self-harm, and child abuse, distinguishing past from ongoing risk. We further propose an Alert-Confirm evaluation protocol that distinguishes early warning signals (Alert) from turns where a specific crisis becomes explicitly identifiable (Confirm), reflecting the clinical need to intervene before risk becomes explicit. Experiments show that identifying when risk emerges is much harder than recognizing that it exists: models achieve only mid-40% to high-60% Micro F1. Additionally, we release a synthetic training corpus and a 32B-parameter model that substantially outperforms existing open-source models and achieves competitive or superior results against proprietary models across turn-level, dialogue-level, and confirm-only evaluation settings.

Diffusion-based action generation has become a foundational component of embodied AI, but its reliance on visual conditioning leaves deployed visuomotor policies vulnerable to adversarial manipulation. Most prior attacks focus on disruption: they perturb the observation stream to reduce task success or induce erratic behavior. We study a stronger threat, Test-time Adversarial Takeover (TAKO), in which an attacker obtains a real-time steering interface over a frozen robot policy and turns it into a remotely piloted instrument. TAKO learns a small vocabulary of reusable universal patches through differentiable diffusion inference; at test time, the attacker switches among these patches in the camera stream to compose attacker-chosen trajectories. This works because the perturbation acts on the visual conditioning pathway, where the induced bias can persist through iterative generative inference. We further show that the natural targeted baseline, target-policy matching, fails because the victim policy cannot reliably supervise itself on out-of-distribution target shifts. Across four tasks (2D manipulation, simulated aerial delivery, simulated ground navigation, and physical-world ground navigation), two visual encoders (ResNet-18 and EfficientNet-B0 + Transformer), and three generative inference families (DDPM, DDIM, and flow matching), human operators achieve 100\% takeover success on attacker-defined objectives in every evaluated setting. The project page is available at https://tako-attack.github.io.

As large language models (LLMs) continue to scale, it becomes increasingly challenging to grow model capacity under fixed computation budgets. We propose Path-Aligned Decompression Distillation (PADD), a framework for distilling knowledge from dense teachers without explicit routing into mixture-of-experts (MoE) students while learning high-quality routing policies. PADD organizes knowledge distillation into four stages in two phases: an initialization phase (Stage I) that builds diverse functionality in the student's experts through teacher neuron clustering and student-expert warmup, and a training phase (Stages II--IV) that integrates online adaptive distillation, path-refined policy optimization, and reward-augmented load balancing in a single training pipeline. Experiments on mathematical reasoning benchmarks demonstrate that PADD yields substantial gains over strong baselines at the same inference cost and that the MoE student can match or surpass its dense teacher. They also demonstrate effective teacher-to-student knowledge distillation and stable routing behavior.

Simulation has become an essential tool for evaluating and improving vision-language-action (VLA) policies, offering scalable, reproducible, and controllable alternatives to costly real-world robot evaluation. Recent simulation benchmarks have made substantial progress on realism and diversity, yet these platforms have not been widely adopted as reliable proxies for real-world policy evaluation. In this work, we investigate this issue through the lens of sim-and-real correlation. We conduct a systematic study across multiple simulation platforms, VLA policies, tasks, and perturbation factors, measuring whether simulated evaluation preserves real-world conclusions in terms of policy ranking consistency, performance correlation, and perturbation-wise failure patterns. This analysis allows us to characterize the limitations of existing simulators and identify what kinds of simulation signals are more aligned with real-world deployment. We further examine how users should exploit simulation for policy improvement, including when simulator-based finetuning is beneficial and how the amount of post-training data affects sim-and-real alignment. Overall, our work provides a unified framework for measuring, interpreting, and improving the usefulness of simulation for VLA policies, offering guidance both for simulator designers and for practitioners who use simulation as part of the policy development pipeline.

Reconstructing printable 3D models from Mars rover imagery is challenging because Martian terrain is low-texture, irregular, and partially observed. We evaluate a pipeline that estimates stereo depth from NASA Curiosity images, completes geometry, and exports watertight OBJ meshes. On Middlebury, RAFT-Stereo outperforms semi-global block matching (SGBM), reducing disparity MAE from 3.22px to 0.73px and increasing valid prediction coverage from 76.3% to 100.0%. On Curiosity imagery, however, RAFT's denser disparities show weaker edge alignment and higher photometric reprojection error, suggesting that benchmark accuracy does not directly transfer to Martian terrain reconstruction. Geometry completion demonstrates a tradeoff between local fidelity and global connectivity. We find that alpha shapes preserve accurate but fragmented structure, Poisson reconstruction produces more coherent meshes but adds unsupported surfaces, and a deterministic diffusion-fill baseline is intermediate but sensitive to stereo quality. Overall, standard stereo and completion methods can produce printable approximations of Martian terrain, but reliable reconstruction requires stronger domain-specific validation.

World Action Models (WAMs) have emerged as a new powerful paradigm for embodied intelligence, learning action-relevant visual dynamics that significantly enhance generalization and robustness. However, existing WAMs still struggle with task-relevant memory in long-horizon robotic manipulation. To address this, we present HiMem-WAM, a Hierarchical Memory-Gated WAM that integrates motion-centric latent actions, high-level skill latents, and boundary-triggered memory updates. Specifically, we develop a hierarchical latent action framework that jointly learns low-level motion and high-level skill latents, providing structured temporal abstraction. Meanwhile, a boundary-aware memory gate writes compact task states at predicted skill transitions, enabling causal inference without test-time generation of future video or optical flow estimation. Evaluated on LIBERO, LIBERO-PLUS, RMBench and real-world tasks, HiMem-WAM shows that hierarchical latents improve robustness under deployment perturbations, and the memory module substantially benefits memory-dependent long-horizon manipulation.

Object navigation requires an agent to locate a target in an unknown environment through visual observations. Existing methods typically rely on open-vocabulary detectors or vision-language models (VLMs) to answer where to search, but often overlook what not to trust - which semantic cues are unreliable. Open-vocabulary perception is prone to systematic misleading evidence: false positives, outdated static priors, and repeated failed exploration due to lack of embodied verification, which contaminates mapping and decision-making. Such errors are rooted in structured object relations in real-world scenes. To address this, we propose DB-Nav, a framework that reshapes the search space via dual relational biases. It factorizes target-centric relations into an Activation Bias (propagates contextual evidence) and an Inhibition Bias (suppresses unreliable regions via perceptual confusion and action-level falsification). These biases are unified into a Relational Activation-Inhibition Exploration Graph that modulates frontier exploration values using online observations and failed accesses. Experiments on ObjectNav benchmarks show that DB-Nav significantly outperforms existing methods in success rate (SR) and Success weighted by Path Length (SPL), offering a lightweight, interpretable, and robust navigation framework without costly online VLM reasoning.

Humanoid whole-body control has made significant progress in recent years, yet existing approaches remain limited to few-skill policies with heavy reward engineering, or motion trackers that are difficult to extend to new input modalities. We argue that the key to general-purpose humanoid control is to build a scalable brain, a module capable of reasoning with diverse conditioning modalities, atop a reactive motion tracking cerebellum, mirroring the hierarchical structure of biological motor systems. Two challenges arise in realizing this vision: acquiring a vast amount of high-quality data to achieve general purpose control, and equipping the generator with the capability to condition on compositional, extensible multi-modal inputs. We present OMG, which addresses these challenges with a meticulous data curation, filtering and labeling pipeline, as well as a diffusion-based motion generation backbone that conditions on language, audio, and human reference motions. Extensive experiments validate OMG as an omni-modal whole-body controller exhibiting state-of-the-art performance, model scaling behavior and efficient adaptation to new distributions and modalities, marking a concrete step toward foundation models for humanoid robots.

Large Language Models (LLMs) in multi-turn interactions maintain evolving context rather than generating isolated responses, making them vulnerable to prompt-injection and context-poisoning attacks in which locally plausible adversarial fragments gradually distort reasoning trajectories. Existing defenses mainly filter individual outputs and often ignore context evolution across turns, leaving long-horizon reasoning exposed. Although the Model Context Protocol (MCP) standardizes context exchange and tool invocation, it functions as a passive routing layer and does not enforce contextual stability. To address these limitations, we introduce the Game-Theoretic Secure Model Context Protocol (GT-MCP), a controller-driven multi-agent method that treats context management as a closed-loop dynamical process. GT-MCP coordinates three heterogeneous LLM agents and selects outputs through a trust function that jointly evaluates causal consistency against a validated context graph, semantic agreement among agents, and distributional drift over time. When instability is detected, a rollback-based self-healing mechanism restores the validated context and prevents unsupported fragments from propagating. Empirical evaluation over 500 interaction turns under an adaptive adversarial threat model shows that contextual drift remains bounded in 99.6% of turns, with recovery required in only 0.4%. Per-turn utility remains tightly concentrated, with median = -0.19, P05 = -0.72, and P95 = 0.30; severe degradation below -1 occurs in only 0.4% of cases, and no injection attempt succeeds at the controller level. Selected outputs maintain stable win rates above 98%, and computational overhead remains predictable, with latency per token = 1.63e-3 s.

Neural combinatorial optimization (NCO) trains autoregressive policies to solve routing problems. The standard training algorithm, REINFORCE with a rollout baseline, requires maintaining and periodically updating a frozen copy of the policy for variance reduction. This baseline introduces a structural vulnerability: on harder instances, a poor baseline produces noisy gradient estimates that can destabilize training. We evaluate Group Relative Policy Optimization (GRPO), an algorithm from large language model alignment that eliminates the baseline entirely by normalizing advantages within groups of sampled trajectories. In a controlled comparison of five RL algorithms on TSP and CVRP benchmarks within the RL4CO framework, we find that: (i) GRPO avoids the training collapse observed with REINFORCE on TSP-100, where performance degrades from cost 9.8 to 52.1 immediately after the warmup phase and does not recover under extended training; (ii) at matched gradient updates, GRPO achieves solution quality within 2% of POMO, a strong AM-based multi-start baseline, while requiring no external baseline; and (iii) P3O, a pairwise preference algorithm also from the alignment literature, is competitive on TSP but shows higher variability on CVRP. These results identify GRPO as a promising baseline-free alternative for NCO, particularly in settings where baseline-dependent training becomes fragile.

Spreadsheets and tables are widely used representations for structured data analysis, but effective analysis still requires substantial manual effort and domain expertise. Recent large language model (LLM) agents can automate parts of this process, but they often provide limited transparency into intermediate decisions, rely on implicit assumptions, struggle with multi-table comparison, and repeat similar workflows without adapting to a user's preferences. This paper presents TabClaw, an open-source interactive AI agent for spreadsheet manipulation and table reasoning. Users upload CSV or Excel files and issue natural-language requests; TabClaw clarifies ambiguous intent, exposes an editable execution plan, streams a ReAct-style tool-using analysis loop, dispatches specialist agents for parallel multi-table reasoning, and synthesizes findings with explicit consensus and uncertainty markers. Beyond one-off analysis, TabClaw records completed workflows, extracts persistent user memory, distills reusable skills from repeated tool-use patterns, supports package-style skill import, and upgrades skills from negative feedback. Experiments on spreadsheet manipulation and table reasoning benchmarks show that TabClaw improves executable task completion and reasoning performance while preserving an inspectable user workflow. This paper shows how TabClaw turns spreadsheets and tables into inspectable analytical workflows while gradually personalizing itself to recurring data-analysis tasks. Our code is available.

LLM-as-judge is the default instrument for evaluating conversational agents, yet its reliability is almost always reported as agreement with human ratings, not recall of real defects. We study a deployed multi-turn food-and-beverage ordering agent and measure how many genuine quality problems its built-in LLM judge catches, using exhaustive human transcript review as ground truth. Across three batches the judge surfaces well under a quarter of human-confirmed systematic problems -- 2 of 9 patterns (22%) in one batch, and its operational gate flagged zero of 100 rounds in a batch where humans confirmed 23 distinct defects and 7 new cross-cutting patterns. Our blind-spot taxonomy shows the failure is structured, not random: the judge catches turn-local issues (a fabricated statistic, a wrong language) but misses cross-turn state issues (confirm-gate lockout, cart hallucination, escalation lockout, stale referents). The mechanism: the scoring rubric exposes only three coarse axes (intent, brand-voice, personalization) and has no category for the behavioural dimensions -- state-tracking, guardrails, recovery -- where most defects cluster. The failure is routing, not perception: 113 of 114 rounds whose raw judge note describes a confirm-gate or cart-state defect are scored "brand voice", and none reach an operational failure -- the gate is wired to hangs and hard assertions, not the rubric -- so the 0% is a routing-and-wiring failure, not blindness. The consequence for prevalence estimation is sharp: when the apparent defect rate is zero the Rogan-Gladen correction degenerates -- no signal can recover the true rate -- while where the gate reports a nonzero rate the same estimator implies a 3-6x undercount under our measured sensitivity. For production multi-turn agents, automated judging is a regression floor, not a substitute for human review.

Evaluating reasoning quality in multi-agent LLM systems is challenging, especially for open-ended tasks without reference answers. We investigate whether intrinsic confidence signals, token-level log-probabilities from decoding, can predict reasoning quality as assessed by LLM-as-judge evaluation. Using a debate-based essay scoring framework, we compare confidence proxies against rubric-based judge scores across two ASAP essay sets. We find that early-token confidence, particularly within the first few generated tokens, is consistently the strongest predictor of reasoning quality, outperforming full-sequence statistics. Analysis of log-probability trajectories shows that the opening phase of generation is the most heterogeneous and therefore most informative. We also observe a systematic asymmetry between agent roles, with stronger alignment between confidence and quality for supportive reasoning than for adversarial critique. These results suggest that early decoding dynamics provide a lightweight and effective signal for estimating reasoning reliability in multi-agent LLM systems.

Fine-tuning vision-language-action (VLA) policies for long-horizon manipulation still relies heavily on behavior cloning, which requires costly high-quality demonstrations and keeps policies near the demonstration distribution. Reward models can reduce this dependence by reweighting demonstrations and providing dense supervision for on-robot reinforcement learning (RL), but they must be dense, accurate, and general. Existing methods fall short: task-specific stage-aware models are accurate but require per-task annotations, while general vision-language-model (VLM) reward models are broadly applicable but too coarse for fine-grained long-horizon progress. We introduce RM, a multi-task stage-aware reward model that combines an action-primitive-based stage estimator with a multi-gate Mixture-of-Experts (MMoE) value head to produce dense per-step rewards across manipulation tasks. Building on RM, we further propose SPIRAL (Self-Policy Improvement via Reward-Aligned Learning), an on-policy reward-guided framework that improves VLA policies from cheap autonomous rollouts. On a 10-task benchmark, RM reduces value-estimation MSE by 80% over the strongest baselines; when used in SPIRAL, it improves task success from around 50% to near-perfect performance on Folding Shorts (58% to 100%) and Cleaning Whiteboard (50% to 90%), showing that high-quality dense rewards are key to a stable robot data flywheel. Project website: https://qianzhong-chen.github.io/sarm2.github.io/.

When LLM agents are coerced into covertly encoding sensitive data (Base64, ROT13, acrostic, synonym chains, and beyond), the resulting outputs evade output-side detection but the underlying computation does not. Across nine encoding families and eight models from five architecture families, that computation is supported by a shared low-dimensional encoding subspace in the residual stream. A logistic-regression probe trained on eight encoding families recovers the held-out ninth at AUC 0.975-1.000, reading the computation rather than surface features. The same direction exhibits a second mechanistic signature at the planning token, flipping polarity to activate positively when the model will simulate the encoding inline and negatively when it will outsource it to a tool call, distinguishing two execution strategies before the encoded text exists. We build MIRAGE (Model-Internal Readout of Agentic Generation Exfiltration), a two-channel real-time monitor exploiting both signals. On 126 agentic exfiltration scenarios, it reaches AUC = 0.918, substantially outperforming output-only detection (AUC = 0.518). Monitor performance is fundamentally a property of the host model's geometry: benign-encoding false-positive rate ranges from 0% on Qwen-7B to 100% on Phi-3.5, revealing that the probe faithfully reads whether a model's geometry separates covert from overt encoding. Across all tested adversarial budgets, every attack suppressing the subspace also destroyed encoding fidelity, reported as an empirical regularity on the evaluated budgets, not a structural impossibility claim.

Language-agent "memory palace" systems anchor each memory to a world coordinate, on the intuition that geometry adds something text cannot. We make that intuition testable and report three results. First, the memory-palace default of folding spatial proximity into a linear blend beside recency and importance does not help and can hurt: in a pre-registered recall experiment the shipped blend fails its own frozen test (mean Delta-Hit@5 -0.0375, Wilcoxon p=0.306), sitting at a position-blind baseline, while a geometry-led weighting wins decisively (+0.3208, p<10^-15): geometry must lead recall when the query regime is spatial. Second, memory recall and visibility must be separated: recall is occlusion-blind by design (you correctly remember the next room behind a wall), while visibility is a perception predicate over stored geometry that the live system never computed. A one-line ray-versus-voxel digital differential analyzer (DDA), re-pointed from the gaze ray the agent already casts, supplies it: text and the live FoV cone both score 0.000 on 849 behind-wall targets while cone-plus-DDA reaches 0.982 (exact McNemar p<10^-6); coordinate recall separately resolves near-duplicate locations a cosine null cannot (1.000 vs 0.533, n=150). Third, the visibility predicate is confirmed live under a git-committed pre-registration (SPMEM-OCC-LIVE-v1: eight scripted worlds, automated oracle scoring, 96 behind-wall targets, false-visible 1.000->0.000, pooled exact McNemar p=2.5x10^-29), a run that surfaced and fixed a real relay anchor defect. We concede that occlusion-needs-geometry is near-tautological; the contribution is the measurement and isolation, separating what spatial memory must store from how it is read. These pilots power a frozen confirmatory study (SPMEM-ZERO-REAL-PREREG-v1); the full human-authored multi-world study with blind raters remains future work.

When large language models generate from retrieved or augmented contexts, conflicts between external context and parametric priors remain a central reliability bottleneck. Existing contrastive decoding methods follow a \emph{context-aware} paradigm that unilaterally amplifies context over parametric priors, overwriting correct priors when the context is erroneous. We generalize this to the \textbf{conflict-aware} paradigm that dynamically allocates authority between prior and context based on conflict signals, rather than presupposing context trustworthiness. We show that the affine combination of prior and context logits yields a \textbf{power family} with an inherent \textbf{regime asymmetry}: extrapolation amplifies errors unboundedly when the prior is correct, interpolation under-corrects when the context is correct, and no static regime covers both. Existing contrastive decoding methods are instances of this family, mostly extrapolative. To evaluate both conflict directions, we propose TriState-Bench, a model-aware evaluation protocol that calibrates per-model prior knowledge to measure three conflict states: correction, resistance, and agreement. To resolve the asymmetry, we propose Adaptive Regime Routing (ARR), which routes between regimes at each step, lifting resistance EM from below 6 to 16--33 without sacrificing correction or agreement. Our code is available at https://github.com/keith-Jiang/conflict-aware-decoding.

Multi-agent debate systems are typically evaluated only on whether the final answer is correct, overlooking the quality of the intermediate reasoning that debate is designed to produce. This paper studies the relationship between three signals in multi-agent debate: token-level log-probability distributions over reasoning tokens, LLM-as-judge rubric scores assigned to those tokens, and final task accuracy. We examine whether internal confidence signals predict externally evaluated reasoning quality, and whether either signal aligns with task correctness, across three domains: rubric-based scoring, mathematical reasoning, and factual question answering. Our framework pairs a two-agent debate architecture -- a Constructor and an Auditor -- with an LLM-as-judge that scores each agent's reasoning along instruction following, justification quality, and evidence grounding, together with a critical-failure flag. Experiments in the rubric-scoring domain reveal a consistent four-phase confidence trajectory and a substantial role asymmetry: confidence aligns with judged reasoning quality roughly twice as strongly for the Constructor as for the Auditor, and confidence-based detection of critical reasoning failures is markedly more reliable for the Constructor (AUROC 0.804) than for the Auditor (0.634). These findings motivate the broader cross-domain investigation proposed in this paper.

Perceptive bipedal locomotion over sparse terrain remains a difficult challenge: model-based methods are precise but brittle to uncertainty, while model-free methods are robust but struggle to discover the precise, constrained motions required for safety-critical locomotion where small errors can cause catastrophic failures. We propose a model-assisted reinforcement learning (RL) framework that combines both perspectives in three steps: (1) generate a safe reference trajectory using simplified models; (2) train a privileged teacher policy guided by a control Lyapunov function (CLF) reward built around the safe reference trajectory; and (3) distill the teacher into a vision-based student policy. We show that this model-assistance procedure produces physically grounded locomotion, improving sample efficiency, reducing the need for a complex learning curriculum, and achieving smoother locomotion behavior alongside stepping stone performance comparable to model-free baselines. We validate our approach in simulation and demonstrate successful deployment on a Unitree G1 humanoid robot navigating sparse footholds with lateral constraints.

Evaluating Knowledge Graph Completion (KGC) models remains challenging because standard assessment relies on isolated rank-based metrics such as MRR, Hits$@$k, and Mean Rank, which often produce conflicting model orderings across datasets. A model that leads on MRR may trail on Hits@1, and strong performance on one dataset may not generalize to another. This fragmentation hinders comparison, enables selective reporting, and obscures real progress. We reframe KGC evaluation as a Multi-Criteria Decision-Making (MCDM) problem and present a meta-analysis of seven aggregators across five tests: consistency, cross-dataset stability, metric independence, robustness under noise, and generalizability. Each test is averaged over leave-one-model-out (LOMO) and leave-one-group-out (LOGO) removals so that reliability reflects aggregator behavior across diverse model subsets. Across tail $(h,r,?)$ and relation $(h,?,t)$ prediction, Pareto-optimal analysis identifies Z-score as the most balanced aggregator, which ranks DualE highest for tail prediction and FMS (Flow-Modulated Scoring) highest for relation prediction. A test-sensitivity analysis using the same removals shows that consistency and stability are largely removal-invariant, while generalizability and independence are the most sensitive. The framework resolves evaluation inconsistencies and offers evidence-based guidance for aggregator selection and model benchmarking in KGC.

This paper presents AuditBench, a new benchmark dataset for evaluating the capabilities of LLMs at investigating security-related system audit logs. We design and use this benchmark to explore the performance of LLMs on four log-investigation tasks that incident response teams commonly perform, ranging from triaging alerts generated by detectors to identifying persistence mechanisms on compromised systems. AuditBench consists of system audit logs collected from Linux and Windows machines, and spans over 50 different security investigation scenarios, including both malicious and benign activity. Using our benchmark, we evaluate and analyze the performance of five frontier LLMs at analyzing audit logs for attack investigations. Our analysis illuminates how LLM performance and error profiles vary according to different design choices, such as differences in model size, data representation, prompt construction, and specific investigation tasks. Additionally, we characterize the quality of the explanations produced by LLMs and the types of errors that models make across our benchmark. Collectively, our work provides a foundation for assessing the capabilities of LLMs for investigating security logs, novel insights for practitioners using LLMs in security operations, and important directions for future research.

For natural human-robot interaction, a robot must understand human intent expressed not only through language but also through nonverbal signals such as gestures and gaze. However, current robot policies rely on language instructions as the sole interface for conveying intent, leaving nonverbal signals unused and placing the full burden of communication. In this work, we present EDITH, a robot framework that captures the human's nonverbal signals through continuous streams of first-person view and gaze from smart glasses, and uses them alongside language instructions as inputs to the robot policy. Our hardware system streams the human's first-person view, gaze, and speech to the robot in real time, transcribing the speech into language instructions. To handle these rich but noisy signals, we design a hierarchical policy in which a high-level policy infers the human's intent and produces a sequence of subtasks, where each subtask is represented as a fine-grained instruction paired with a keyframe that grounds the intent in the scene (e.g., the frame where the human points at the target object). A low-level policy then executes these subtasks. In our experiments on human-robot interactive tasks, EDITH enables the robot to act on the human's nonverbal signals even when intent is expressed only briefly, and significantly reduces user effort to convey intent compared to using language instructions alone. Visit our project page for source code and real-robot demo videos.

Deploying legged robots in extra-terrestrial environments includes many challenges due to complex terrain interactions, energy, and thermal constraints. For effective mechanical design of a lunar exploration quadrupedal robot, careful consideration of motor torques, energy expenditure, and cost of transport is required. The lunar surface is composed of granular regolith, which impacts the locomotion of legged robots and their performance. Locomotion algorithms trained with rigid contact assumptions are also ineffective when applied to environments with soft contacts, such as granular surfaces, which can result in instability and poor tracking. In this report, the physical modelling of the granular lunar surface-robot foot contacts is applied to a simulation environment with locomotion trained using Reinforcement Learning. A comparison is conducted between the policy trained on rigid contact and soft contact environments, analysing the gait and locomotion performance metrics. The analysis demonstrates that soft contacts simulating regolith surfaces pose additional challenges for Reinforcement Learning based training, result in a qualitatively different gait, and increase the overall energy expenditure.

Hierarchical vision-language-action (Hi-VLA) systems have emerged as a promising paradigm for complex robot manipulation, by using high-level VLM planners to decompose tasks into language subgoals executed by low-level VLA controllers. Despite recent empirical progress, there is a lack of unified design principles for these systems: existing Hi-VLA systems differ in how they choose and connect planners, controllers, mechanisms to switch between the two, and how observations and memory are represented in the planner. In this paper, we present a systematic study of Hi-VLA design for robot manipulation. We unify representative Hi-VLA agents under an options-style control framework and benchmark core design choices across short-horizon, long-horizon, and reasoning-intensive tasks. Our analysis distills practical principles for building Hi-VLA systems, showing how model choices and interface mechanisms jointly shape performance. Applying these principles yields a substantially stronger system than either flat VLA control or a naively designed hierarchy, across experiments both in simulation and on a real ALOHA robot. Overall, our results provide a foundation for building more capable, robust, and principled hierarchical VLA agents. More information and video at jiahenghu.github.io/hi-vla.

We performed a 2D finite-element electrothermal computational study of thermoelectric effects and filamentary electronic transport in Ge$_2$Sb$_2$Te$_5$ mushroom phase change memory cells during Reset and Set operations, accounting for spatial activation energy variations in amorphous Ge$_2$Sb$_2$Te$_5$ and phase-change dynamics. Reset operations with current going from the top electrode to the narrow 4 nm bottom electrode require $\sim$3x less energy and power, and $\sim$2x lower current to achieve the same Reset resistance, compared to the opposite polarity, due to thermoelectric effects. Filamentary conduction, electrical breakdown, thermal runaway, and local crystallization of amorphous Ge$_2$Sb$_2$Te$_5$ depend on current polarity and thermal boundary conditions, and determine the location, shape, and volume of the programming region, which may be significantly smaller than the semi-cylindrical mushroom region. The programming volume does not scale with contact dimensions larger than 10 nm. Larger contact areas introduce increased device-to-device and cycle-to-cycle variability due to filamentary conduction but are expected to lead to higher reliability and endurance.

Cryo-electron tomography (cryoET) has emerged as a powerful tool in structural and cellular biology by enabling direct visualization of macromolecular structures within intact cells, thereby linking molecular architecture to cellular organization in a native context. Realizing the full potential of cryoET, however, increasingly depends on advances in computational analysis, particularly machine learning (ML), to interpret its complex and information-rich data. Despite rapid progress, ML development for cryoET remains bottlenecked by the lack of standardized, well-annotated benchmarks. Existing evaluations are typically small, task-specific, and are assembled in isolation, limiting robust comparisons across methods. Here, we present POPSICLE, a benchmark suite for cryoET segmentation and macromolecular localization built from the CryoET Data Portal - an open, ML-ready repository of tomographic data, metadata, and annotations. POPSICLE spans eukaryotic and prokaryotic systems, both purified and fully in situ samples, and dense voxel-wise segmentation as well as sparse localization tasks. Built on a living data resource, it can expand as new datasets and annotations become available. Baseline experiments reveal substantial variation in model rankings across tasks, underscoring the need for benchmarks tailored to the unique characteristics of cryoET rather than evaluation practices adapted from adjacent biomedical imaging domains. POPSICLE thus provides an open and extensible foundation for reproducible ML evaluation in cryoET.

While Large Language Models (LLMs) have achieved near-perfect performance in \emph{solving} high-school mathematics, their ability to \emph{evaluate} the diverse reasoning processes of real human students remains under-examined. To bridge this gap, we introduce \textbf{RealMath-Eval}, a rigorously annotated benchmark of 224 real-world exam responses from high schools. Our initial evaluation reveals that even state-of-the-art LLM judges struggle significantly on this task, exhibiting a high Mean Squared Error ($\sim$2.96) against expert human grading. To probe a plausible explanation, we contrast this performance with a control setting where the same judges evaluate synthetic LLM-generated solutions. We identify a stark ``Evaluation Gap'': judges are considerably more accurate and consistent on synthetic text (MSE $\sim$1.17) but struggle to generalize to authentic student reasoning. Through semantic embedding analysis, we find that synthetic errors suffer from a ``structural collapse'' into predictable, low-dimensional linear subspaces, whereas human errors form a more diverse error space. Furthermore, generative probability probes suggest that human reasoning involves significantly higher information-theoretic surprisal, indicating that student reasoning transitions are more out-of-distribution for current models. Finally, we find that surface-level style transfer fails to close this gap. Our findings suggest that current LLM evaluation pipelines relying heavily on synthetic data may not adequately capture the diversity of authentic student mathematical reasoning.

We introduce Yielding Universal Bidigital Interface (YUBI), a finger-aligned gripper designed to enable intuitive, ergonomic, and scalable data collection for bimanual dexterous manipulation. While handheld data collection systems such as Universal Manipulation Interface (UMI) enable affordable data collection, their bulky pistol-grip designs can pose ergonomic and usability challenges for fine-grained, dexterous manipulation tasks. To address this, YUBI presents a distinct design principle: yielding, finger-driven actuation that directly maps human finger movements to gripper jaw motion. Using the YUBI devices, we set up a data collection system with integrated VR-based 6 DoF tracking of the gripper, ensuring high-fidelity trajectory data acquisition. We curate a UMI-based dataset of unprecedented scale: 8,434 hours across 1.20M episodes and 119 tasks. Experiments show that YUBI offers advantages over the UMI gripper in versatility for complex bimanual tasks, dexterity, and operational efficiency. A single policy trained on the YUBI dataset transfers across multiple bimanual robots (UR, Franka, and ELEY) simply by mounting the gripper on each platform, confirming that the collected data are directly executable as policy supervision. We release the gripper hardware, data-collection software, and dataset as one integrated stack, offering the open community a reproducible path to large-scale data acquisition for advancing robotic foundation models.

We study whether demonstration-curation metrics that detect defective training episodes also improve the downstream behavior-cloning policy that trains on the curated data. On a contact-rich LIBERO pick-and-place benchmark with a controlled structural defect (early gripper release during the carry phase), we find that the two quantities are sharply decoupled. The metric with the highest defect-detection AUROC (0.804) produces the worst curated policy (13.3% task success), while a metric with a substantially lower AUROC (0.638) produces a policy that nearly matches the oracle trained on ground-truth clean data (90.0% vs. 93.3%). We further show that five of the seven metrics we evaluate exploit episode length as a trivial proxy for the defect label, a confound that inflates reported AUROCs to near-perfect values and disappears once episode length is controlled. Across all conditions, the contaminated baseline succeeds on only 3.3% of rollouts, and the two best curation methods close this to within 3 percentage points of the 93.3% oracle ceiling. Our results argue that curation methods should be evaluated by the policy they produce, not the defects they flag, and that any curation benchmark must control for episode length before reporting detection accuracy. We release the testbed, all metric implementations, and the evaluation pipeline.

Attributing a synthetic utterance to its originating system remains an open challenge: closed-set models fail to reject unseen synthesizers and produce overconfident predictions. To address this, we propose a dual-branch gated fusion framework that pairs XLSR-53 with CORES, a 66-dimensional descriptor that, unlike prior Linear Filter Bank (LFB)-only work, spans cepstral, oscillatory, rhythmic, energy, and spectral dimensions to capture complementary synthesis artifacts. Our analysis shows XLSR-53 remains discriminative in-domain (ID) while CORES generalizes stably under distribution shift (OOD), yet their naive concatenation fails due to SSL representational imbalance. To resolve this, an input-conditioned gate adaptively weights each branch under joint training with cross-entropy, an energy margin loss for ID/OOD separation, and a gate diversity term. On the MLAAD benchmark, our system achieves 97.6\% ID accuracy, 4.9\% EERc, and an 83.5\% relative FPR95 reduction over the Interspeech 2025 baseline.

Demand for older-adult and patient care is growing rapidly as populations age worldwide. Foundation models are increasingly being integrated into robots and interactive agents, with the promise of more flexible communication and personalized assistance. However, care settings require reliable and workflow-compatible systems with accountable human oversight, and it remains unclear whether current embodied systems can translate technical advances into clinical impact. This Perspective synthesizes foundation model-based care robots across three areas: design features, user experience, and evidence for care-related outcomes. Current systems most commonly use foundation models as conversational and reasoning layers within voice-centered socially assistive embodiments, while multimodal grounding and physical autonomy remain limited. Empirical evaluations report positive usability and engagement benefits, but reliability failures persist across the interaction pipeline such as hallucinations and conversational breakdowns. Evidence for care impact remains concentrated in proximal outcomes such as cognitive engagement and participation, with limited evidence for validated clinical or care-related changes. We argue that future research should transition toward care-specific evaluation standards, accountable autonomy, and integration into care workflows to support more responsive and responsible care technologies.

We formulate a multi-agent area coverage control problem as a two-player zero-sum game between two agent groups with conflicting goals. Conventional coverage control allocates resources based on an environmental risk density field. In contrast, we generalize this metric by allowing a second group of adversarial agents to generate the spatial risk field. Coupled agent dynamics are linked through the area coverage metric, which functions as the game reward. This framework induces coupled gradient-descent-ascent controllers for the groups. Analysis of a low-dimensional case reveals a Hopf bifurcation dictated by the ratio of the groups' control gains. In the regime dominated by adversarial agents, the system is driven into a periodic chase-evasion cycle. In the regime dominated by ordinary agents, the system converges to a fixed configuration. Numerical simulations validate these theoretical insights. Finally, we characterize the Nash equilibrium conditions. Under this equilibrium, ordinary agents converge to a generalized centroidal Voronoi tessellation, whereas adversarial agents settle at their corresponding equilibrium centroids.

Hallucination detection in large language and vision-language models is increasingly framed as selective prediction, where a detector assigns a confidence score and abstains when confidence is low. Unsupervised sampling detectors (Semantic Entropy, EigenScore) avoid labels but plateau in quality, while supervised probes (SAPLMA) attain stronger in-distribution scores yet degrade sharply when calibration labels are scarce. We recover the response manifold of an LLM as the density ridge of a kernel density estimate built on a six-dimensional kinematic feature map of hidden state generation trajectories. A test generation is scored by the negated Euclidean distance from its projected feature point to the nearest ridge vertex, yielding a low-dimensional geometric skeleton of the stochastic output distribution. We evaluate against Semantic Entropy, SAR, EigenScore, SAPLMA, and log-probability on seven QA benchmarks (HaluEval-QA, TriviaQA, GSM8K, POPE, ScienceQA, A-OKVQA) using nine text and vision LLMs in a deliberately label-scarce protocol ($n_{\text{cal}}{=}200$ queries, $N{=}5$ generations). Our ridge-based score beats on AUROC with 5-20 points gain, while demonstrating tempered degradation under calibration-label scarcity.

Submodular set functions, characterized by the diminishing-returns property, provide a unifying combinatorial framework for many subset-selection problems in decision and control. Although exact maximization is NP-hard in general, the structural properties of submodular functions enable simple greedy algorithms that achieve constant-factor approximation guarantees for monotone objectives, with randomized greedy-based variants extending such guarantees to the non-monotone case. This survey reviews the theory, algorithms, and applications of submodular optimization with a focus on systems and control. We cover the structural properties of submodular functions, including curvature and the submodularity ratio, the constraint families that arise in practice (matroids, knapsack, and $p$-systems), and the main approximation algorithms for monotone and non-monotone submodular maximization, with up-to-date approximation ratios and hardness results. We then survey applications across sensor scheduling, multi-agent coordination, robust submodular optimization, leader-follower systems, distributed submodular optimization, game theory, system theory, resource allocation, social networks, and informative path planning. The survey emphasizes practically implementable greedy-based algorithms and instance-dependent refinements via curvature and the submodularity ratio. We close with observations on canonical control-theoretic objectives: certain functionals are submodular (the log-determinant and rank of the controllability Gramian, and the log-determinant of the Kalman filter information matrix), whereas closely related objectives fail to be sub- or supermodular (the steady-state Kalman filter error covariance, and the average control energy obtained from the inverse Gramian). We also highlight the cross-cutting open directions that follow.

We introduce flow control of vision-language-action (VLA) models, a simple and effective way to steer VLA actions in real-time through generic inputs, such as a keyboard. This method can be used out-of-the-box and does not require retraining or fine-tuning VLAs. It enables relatively crude user inputs to steer a VLA to align with user intent. The VLA transforms these inputs into action samples drawn from the VLA expert action distribution learned during training, so that the generated actions are high quality (conformity to the action expert distribution) and high fidelity (reflecting the user's intent). We demonstrate that flow control has many desirable properties: (1) flow control accurately and responsively steers robot actions with user inputs, (2) it is robust to suboptimal user inputs, (3) it enables users to steer VLAs to achieve significantly higher success rates and faster task completion, and (4) fine-tuning a VLA on flow control trajectories improves the autonomous policy. Together, these results provide a simple and intuitive way for users to help steer VLA actions, increasing task performance.

Wildfire detection and monitoring are critical for mitigating fire spread and reducing environmental and infrastructural damage. In this work, we introduce GWFP (Global Wildfire Prevention Dataset), a large-scale, open-source dataset of wildfire images and videos designed to support early fire and smoke detection research. GWFP contains geographically diverse wildfire scenes, including flames, smoke, Waterdog/Fog environmental conditions, Near Infrared (NIR) imagery, Ember, and challenging negative samples collected from real-world scenarios worldwide. To evaluate dataset robustness and cross-domain generalization, we benchmark multiple convolutional and transformer-based architectures across both in-domain and cross-dataset settings. Additionally, we explore lightweight frequency--spatial feature interaction using Hadamard-enhanced residual connections (HTE-ResNet) to analyze representation robustness under domain-shift conditions. Experimental results demonstrate strong cross-dataset generalization and practical utility for real-world wildfire monitoring applications. The dataset and source code will be publicly released upon acceptance.

AI is increasingly used to support scientific peer review, from manuscript screening, reviewer assistance to editorial triage. Although such systems promise to reduce reviewer burden and accelerate publication, their robustness to strategic manipulation remains poorly understood. Here we show that AI-mediated peer review is vulnerable to a simple, low-cost manipulation: superficial rephrasing of the manuscript abstract. Without changing the underlying scientific content and communication, and even without knowledge of the reviewing model, adversarially rewritten abstracts substantially improve AI review outcomes. We see this across disciplines and publication venues, for both human-written and AI-generated papers. Our strongest attack achieves an attack-success-rate of about 38%, increasing acceptance ratings by +1.31 for Gemini 3 Flash reviewers and by +0.88 for GPT 5.4 Mini reviewers on a 10-point scale. When the original AI review suggests 'reject', the success rate rises to more than 50%. This effect extends beyond overall score inflation, increasing review confidence and scores on core scientific criteria such as soundness, significance and perceived contribution. The attack is practical, requiring only about 5 minutes and $1 for a 10-page AI conference submission, and is hard to distinguish from ordinary scientific editing. Inflated AI reviews could bias downstream human decision-making, shifting editorial recommendations from rejection towards acceptance. These findings reveal a general vulnerability in AI-assisted scientific evaluation: when AI-generated review influence editorial decisions, authors may be incentivized to optimize manuscripts for AI judgment rather than scientific merit. Our results suggest that AI tools should not be treated as neutral evaluators in high-stakes peer review without systematic robustness testing, transparent safeguards and careful human oversight.

As recommender systems transition toward agentic, multi-turn conversational interfaces, evaluation paradigms have struggled to keep pace. Current benchmarks often rely on "LLM-as-a-judge" evaluations, which introduce subjectivity, high costs and inconsistency. We present $τ$-Rec, a benchmark for agentic recommender systems that replaces subjective evaluation with verifiable rewards and a reveal-tagged elicitation (RTE) mechanism that controls how task constraints surface during dialogue. By testing agents against structured catalog predicates and employing a pass^k reliability metric, $τ$-Rec provides a systematic test for consistent reasoning. Our evaluation of nine configurations across five model families -- GPT-5.4, Claude Sonnet 4.6, Gemini 2.5 Flash, DeepSeek V4 Flash, Qwen3-32B and GPT-5 mini -- reveals a steep reliability cliff, where even the best model achieves only ~57% at pass^1 and ~38% at pass^4, highlighting a critical gap in current conversational agent deployment. All code and data are publicly available at https://github.com/nbharaths/tau-rec.

Reservoir computers benefit from the inherent complexity of optical phenomena, which provide rich, often nonlinear dynamics. However, training directly on the reservoir's output renders the system prone to overfitting and computationally inefficient during the training phase. In this work, we investigate strategies to mitigate overfitting and reduce computational overhead through output pruning and regularization. We compare loss-minimizing search methods (Equal Search and Branch and Bound) against an output-oriented statistical filtering approach (Variance Filter) and random pruning, highlighting advantages and disadvantages of each approach and the overall importance of informed reservoir output sampling, particularly for a shrinking latent space. We further demonstrate that enforcing readout selection across the full output spectrum improves performance, especially for non-iterative methods. Additionally, we examine L1 and L2 regularization techniques (LASSO and ridge regression), both of which significantly enhance performance on highly nonlinear tasks such as the Spiral Benchmark. While our methods are of general use, results are obtained from and discussed exemplarily for a nonlinear fiber-optical extreme learning machine. Overall, this study provides a deep analysis of the reservoirs' hidden-layer filtering mechanisms and the output-layer training, enabling optimized performance in physical reservoir computing systems.

Pre-training on raw code teaches syntax but provides sparse signal for diverse real-world task formats. While synthetic data has proven transformative for language models, code remains largely unexplored beyond limited quality improvements. We present CodeAlchemy, a synthetic data generation framework that transforms publicly sourced code into semantically-rich training data through 5 strategies: CodeEnhance (quality-aware rewriting), CodeQA (template-based problems), CodeDev (developer tasks), CodeDialogue (multi-turn conversations), and CodeTrace (execution traces). We process 3 corpora across 15 languages to generate 500B+ tokens of synthetic data plus 350B reasoning tokens, orders of magnitude more than prior efforts. CodeTrace instruments and executes 1.3M+ files across 14 languages and 5K libraries, capturing control flow, state tracking, and library knowledge. We introduce DevEval (developer tasks) and TraceEval (execution prediction) benchmarks; frontier models like Claude Sonnet 4.5 achieve only 5.6% exact match on TraceEval, revealing critical gaps in semantic understanding. Our 3B models achieve 83.5% on HumanEval, 63.2% on MBPP, 8.09% win rate on DevEval, and 15.36 ROUGE-2 on TraceEval, outperforming frontier models 10x the size including 27B Gemma-3 and 32B Granite-4.0.

Foundation-model agents are increasingly long-lived systems that remember users across interactions, making memorization an explicit deployment-time function rather than solely a property of model weights. Existing work addresses parametric memorization or audits fixed memory configurations, but does not characterize how memory-design choices jointly shape personalization utility, extraction risk, and deletion fidelity. We study this surface as deployment-time memorization, formulating agent memory as a privacy-utility frontier measured by Personalization Recall (PR) and Adversarial Extraction Rate (AER), and sweeping three memory-design knobs: summarization aggressiveness, retrieval breadth (k), and deletion mode. We further introduce the Forgetting Residue Score (FRS) to quantify whether deleted information remains recoverable from derived memory tiers. On LongMemEval, key-fact summarization reduces canary extraction by 76% on Gemma 3 12B and 64% on GPT-4o-mini while preserving nearly all personalization recall; critically, once content is compressed away, increasing k no longer restores leakage. The same compression, however, induces a deletion-fidelity failure: raw-only deletion leaves derived summary copies recoverable in approximately 20% of instances, and only full-pipeline purge or tombstone redaction drives worst-tier residue to zero. Together, these results establish that persistent agent memory must be evaluated as a first-class memorization mechanism -- assessed by what it helps agents recall, what it makes extractable, and what it can truly erase.

Large language models (LLMs) increasingly participate in emotionally sensitive social conversations, where responses may shift from balanced support toward excessive validation or escalatory alignment. Existing sycophancy research primarily focuses on factual agreement and instruction-following settings, leaving culturally grounded conversational sycophancy underexplored. We introduce BenSyc, the first benchmark for studying conversational sycophancy in Bengali social contexts. Starting from 11,840 Reddit posts and 170k comments collected from communities across Bangladesh and West Bengal, we construct a human-validated benchmark with binary labels and a fine-grained five-level taxonomy spanning Invalidation, Neutral, Support, Validation, and Escalation. We evaluate more than 15 open and proprietary LLMs on conversational alignment classification and response generation tasks. Results show that distinguishing empathetic support from reinforcement-oriented validation remains challenging even for frontier instruction-tuned models: the best system achieves only 61.8 Macro-F1 on binary detection and 61.7 Macro-F1 on five-class classification. In generation settings, several models frequently produce strongly validating or escalatory responses in emotionally charged situations. Our findings highlight substantial variation across model families and conversational behaviors, underscoring the importance of culturally grounded multilingual benchmarks for evaluating socially aligned conversational AI systems.

World-Action Models (WAMs) have emerged as a promising paradigm for embodied control by coupling future visual prediction with action generation. However, most existing WAMs rely on photorealistic future prediction, which incurs high inference latency and makes real-time robot deployment difficult. This motivates a more efficient WAM design that preserves the control benefits of future visual prediction while reducing its inference cost. We introduce Efficient-WAM, a World-Action Model that reduces the cost of future imagination while preserving its control benefit. Efficient-WAM improves inference efficiency via a compact video expert transferred from WAN-2.2-5B, token-sparse video latents, and asymmetric video-action denoising that allocates fewer sampling steps to video than to actions. Instead of optimizing the future branch for visual fidelity, Efficient-WAM treats future video prediction as a compact guidance signal for action generation. Comprehensive experiments on RoboTwin 2.0 and real-world manipulation tasks show that Efficient-WAM maintains strong action performance despite visibly coarse future predictions. While maintaining competitive control capabilities, our 1B-parameter model can reduce per-chunk latency to around 100 ms during physical deployment, achieving a 30x speedup over existing WAMs.

We consider the nonprehensile object transportation task known as the waiter's problem, in which a robot must move an object balanced on a tray from one location to another. In contrast to prior works on the robotic waiter's problem, which make the robot tilt a tray rigidly held by its end effector (EE), we use a tray suspended from the EE by ropes, such that it behaves like a three-dimensional pendulum. Some prior works have actuated the robot so that the EE simulates the behavior of a pendulum, because pendular motion reduces the shear forces acting on the transported objects, minimizing the sliding of rigid objects and sloshing in containers of liquid. In contrast, our use of a real hanging tray allows us to obtain the benefits of pendular motion while only actuating a 3 degree-of-freedom (DOF) mobile base, rather than requiring a full 6-DOF manipulator arm. Our experiments in simulation and on real hardware show that the hanging tray substantially reduces both sliding and sloshing compared to a static, rigidly-grasped tray. Furthermore, we integrate the hanging tray into an interactive robot waiter demonstration, which uses computer vision to identify people with a raised hand and visual servoing to steer toward them and allow them to access the tray.

We present GHOST, a framework for learning visuomotor manipulation policies that generalize beyond the training distribution. GHOST factorizes control into (i) a high-level policy that predicts the next sub-goal as a distribution over 3D end-effector poses from multi-view RGB-D observations, and (ii) a low-level goal-conditioned controller that executes embodiment-specific actions. To condition image-based policies on 3D goals, we introduce a simple spatial interface that projects predicted goals into the image plane and represents them as end-effector heatmaps. Across a suite of manipulation tasks, this hierarchical factorization consistently improves performance and robustness compared to a flat Diffusion Policy. Further, we show that this hierarchical interface also makes it easy to incorporate human demonstrations without relying on (noisy) action retargeting. As sub-goals are largely embodiment-agnostic, we train the high-level policy on human video to specify how learned skills should be applied and composed, while keeping the low-level policy trained purely on robot data. This hierarchy enables adaptation to novel objects and task variations using a small number of human demonstrations.

We propose a fast and correspondence-free local point cloud registration method that leverages geometric surface structure and reproducing kernel Hilbert space (RKHS) embeddings. The method represents point clouds as continuous functions with point-wise anisotropic kernels that encode local geometry. This formulation improves alignment along surface normals while relaxing alignment along tangential directions. To solve the resulting registration problem, we propose a second-order on-manifold optimization scheme with approximate Riemannian Hessians, achieving a speedup of up to 10x over the first-order solvers used in prior correspondence-free RKHS-based methods. We demonstrate improved frame-to-frame LiDAR and RGB-D tracking accuracy across diverse indoor and outdoor datasets. On a LiDAR tracking registration task in the driving domain, we achieve a reduction of $>55\%$ in both translational and rotational drift in challenging feature-sparse environments. On object registration benchmarks, we show improved robustness over ICP-based methods and further gains when refining global initialization, particularly under moderate misalignment.

Temporal modeling is essential for robotic manipulation, as effective control requires both memory of past interactions and imagination of future states. However, most VLA models rely primarily on the current observation and therefore struggle with long-horizon, temporally dependent tasks. Cognitive science suggests that humans rely on working memory to buffer short-lived context, the hippocampal system to preserve episodic memory of past experience, and internal models to imagine possible future state evolution. Inspired by these mechanisms, we propose MemoryVLA++, a full temporal modeling framework that equips VLA models with memory and imagination for robotic manipulation. A pretrained VLM encodes the current observation into perceptual and cognitive tokens, forming working memory. These tokens query a Perceptual-Cognitive Memory Bank to retrieve relevant historical context. This bank stores low-level details and high-level semantics from past interactions, and is updated through redundancy-aware consolidation. A world model imagines future states in a denoising latent space, and the imagined latents are integrated under memory guidance to form full temporal-aware tokens. The resulting tokens condition a diffusion action expert to predict temporally consistent action sequences. We conduct extensive experiments on 5 simulation benchmarks and 3 categories of real-robot tasks across 3 robots, covering general manipulation, long-horizon temporal tasks, robustness, and generalization. Our method achieves strong performance across Libero, SimplerEnv, Mikasa-Robo, Calvin, Libero-Plus, and diverse real-robot tasks, validating the effectiveness of full temporal modeling with memory and imagination. For example, on real robots, it achieves +9%, +26%, +28% gains on general, memory-dependent, and imagination-dependent tasks. Project Page: https://shihao1895.github.io/MemoryVLA-PP-Web

Training reinforcement learning (RL) policies from scratch is costly: it requires careful reward and environment design, extensive tuning, and substantial computation. Yet many control problems already have a functional but suboptimal policy available as a baseline. This paper proposes a method for embedding such a baseline into the RL training process, simultaneously improving training efficiency relative to from-scratch methods and producing a learning policy that outperforms the baseline. At each step, the method arbitrates between the baseline policy and a trainable learning policy, initially relying strongly on the baseline policy and then progressively transferring agency to the learning policy. By the end of training, the learning policy is a standalone neural network that operates without baseline policy support. The paper formalizes what it means for the baseline policy to be functional: under this policy, the agent reaches a goal set and remains there with high probability. The proposed arbitration mechanism is designed to exploit this property during training, yielding high goal-reaching rates right from the beginning of training. A theoretical analysis provides a formal interpretation of this behavior under stated assumptions and extends it to the final baseline-free regime, where explicit lower bounds are derived for the goal-reaching probability of the standalone learning policy. Empirical results on continuous-control benchmarks show that the proposed method achieves returns that match or exceed those of competitive approaches, while maintaining the highest goal-reaching rates throughout training among the compared methods -- including in the final stage, where the learning policy operates without any baseline support.

Embodied world models have emerged as a pivotal paradigm for visual robotic decision-making and interactive environment simulation. However, conventional embodied frameworks rely on low-dimensional structured action vectors (e.g., joint angles and end-effector poses), which suffer from limited expressive capacity, poor generalization across diverse embodiments, and unnatural dynamic modeling for complex physical interactions. To address these limitations, this paper proposesiMac (Image as Action Control), a novel unified control paradigm that treats raw visual images as native action representations for embodied world models. Departing from traditional explicit kinematic action encoding, iMac formulates continuous visual manipulation as image-based action tokens, which inherently encapsulate spatial motion intentions, interactive geometric constraints and subtle physical dynamics. We construct a dual-branch embodied architecture consisting of an image-action encoder and a dynamic world predictor: the encoder compresses target-driven visual images into compact action embeddings, while the predictor learns environment transition rules conditioned on image actions to achieve high-fidelity future state prediction and closed-loop embodied control. Extensive experiments are conducted on public embodied manipulation benchmarks and real-world robotic scenarios. The results demonstrate that iMac outperforms vector-based action control baselines in prediction accuracy, task success rate and cross-scene generalization ability. Moreover, our image-action design eliminates the reliance on manually defined action spaces, realizing flexible and universal control for heterogeneous embodied agents. This work provides an innovative visual-action perspective for embodied world models, offering a simple yet effective paradigm for scalable robotic perception and manipulation.

World-action models have emerged as a promising paradigm for robot manipulation, jointly modeling visual scene dynamics and actions to inject physical priors into policy learning. However, existing world-action models couple world prediction and action execution at the same temporal resolution, forcing the world branch to model near-term frame variations that are redundant and weakly informative. We posit that strictly binding world prediction and action execution to the same temporal rhythm may underutilize the potential of the video branch for embodied control. Therefore, we propose AHA-WAM, an Asynchronous Horizon-Adaptive World-Action Model built on a dual Diffusion Transformer (DiT) architecture that reorganizes world-action modeling around this temporal asymmetry. AHA-WAM instantiates the video DiT as a low-frequency world planner that maintains rolling key-value memory over past observations and exposes reusable layerwise latent context encoding long-horizon scene evolution, while a high-frequency action DiT executes short action chunks in closed loop by querying this context through layerwise joint attention. To support asynchronous execution, we introduce horizon-adaptive offset training and Observation-Guided Video-Context Routing (OVCR), which together let the action expert exploit long-horizon world context while remaining responsive to real-time execution state without rerunning the video DiT. Experiments on RoboTwin and real-world manipulation tasks show that AHA-WAM achieves state-of-the-art performance without any robot-data pretraining, attaining 92.80% average success on RoboTwin and 78.3% success across 4 real-world tasks, while reaching 24.17 Hz closed-loop control with a 4.59x speedup over Fast-WAM.

Multi-agent code generation offers a promising paradigm for autonomous software development by simulating the human software engineering lifecycle. However, system reliability remains hindered by LLM hallucinations and error propagation across interacting agents. While semantic entropy provides a principled way to quantify uncertainty without ground-truth answers, current methods often rely on costly LLM-driven equivalence checks. In this work, we introduce Fast Adaptive Semantic Entropy (FASE), a novel metric that approximates functional correctness based on the minimum spanning tree of structural and semantic dissimilarity graphs. Evaluations on HumanEval and BigCodeBench demonstrate that FASE outperforms state-of-the-art semantic entropy by LLM entailment, achieving a 25% average improvement in Spearman correlation and a 19% increase in ROCAUC score against Pass@1 from ground-truth test cases when using the Qwen3-Embedding-8B model. Furthermore, by eliminating costly LLM-driven equivalence evaluation, FASE incurs negligible computational overhead, requiring only approximately 0.3% of the runtime cost of traditional semantic entropy approaches. These results position FASE as a practical, cost-effective solution for optimizing uncertainty quantification in real-world multi-agent workflows.

Human hand-object interactions encode functional intent, but direct transfer to robotic hands often fails under morphology, contact, and reachability constraints. We present SynManDex, a synthetic pipeline that uses generated human pre-grasps as affordance-aware proposals and resolves the final contacts with robot-native optimization. SynManDex samples object-conditioned digital human pre-grasps, retargets them to dexterous robotic hand poses, optimizes force-closure contacts on the target embodiment, and admits trajectories that pass checks from each step. The resulting keyframes support both grasp-and-lift demonstrations and various prehensile manipulation tasks such as tea pouring, photo taking, and flute playing, designed via VLM agents. As a result, SynManDex combines high grasp quality (86.4\% grasp stability) with 4.67/5 human-likeness (93.4\%). It achieves 80.7\% successes in simulation and 25/30 (83.3\%) real-robot successes when applied to a 36-DOF bimanual dexterous robotic platform.

Force and tactile sensing are indispensable in contact-rich manipulation. However, force-aware robot learning faces critical challenges due to the incompatible assembly of tactile and force sensors in handheld or wearable devices. To address these limitations, we first introduce AetheRock for gripper-force, vision, and tactile data collection, which is an arm-worn device featuring a modular and easily manufactured visuo-tactile sensor, GelSlim-MiniFab, at the fingertip, a resistive pressure sensor at the human finger contact region, a customized PCB module, and a wearable kit for comfortable and robust collection. Building on this, we propose ForceVT, a representation learning framework that uses force and vision to guide fidelity-agnostic tactile learning, enabling robust inference in any tactile situation. Real-world experiments show that AetheRock achieves qualified data efficiency and that ForceVT effectively alleviates inefficiencies when visuo-tactile sensors exhibit manufacturing and utilization inconsistencies. Overall, our work mitigates the limitations of gripper-force vision-tactile robot learning through innovative hardware design and algorithms.

Advanced scientific simulators expose specialized input languages that turn simulation goals into executable configurations, but learning them can cost domain scientists hours to days. We study simulator setup as a problem of agent-tool interface grounding: what minimal simulator-specific adaptations are needed for an off-the-shelf coding agent to operate real scientific software? Our intuition is that coding agents already know how to navigate files, edit code, run commands, and repair outputs, but they lack the simulator's executable contract: its vocabulary, structural constraints, validation rules, and termination conditions. We introduce SIGA, a Simulator-Interface Grounding Adapter that supplies this contract through retrieval, procedural memory, in-trajectory validation, and validation-enforced termination. We primarily evaluate SIGA on GEOS, an open-source multiphysics simulator used in subsurface science. SIGA produces a complete GEOS deck in about five minutes with TreeSim above 0.90, matching an extended-budget human expert who took about three hours, a roughly 36x wall-clock speedup. On a harder held-out set, grounding raises TreeSim from 0.720 to 0.789, a roughly 10% relative gain over the bare agent, and can reduce the across-seed standard deviation by 16x. Self-evolution further improves SIGA by rewriting adapter contents from prior trajectories, yielding the highest held-out GEOS mean and matching or outperforming the strongest hand-designed configuration. Transfers to OpenFOAM and LAMMPS show that the dominant mechanism shifts by interface: validation matters most when structural completeness is the bottleneck, while memory and retrieval matter most when domain correctness is the bottleneck. These results suggest that lightweight, self-improvable grounding layers can turn general coding agents into practical operators of scientific software.

Neural machine translation for digitally low-resource Indigenous languages is often hindered by extreme data scarcity, prompting reliance on extractive web-scraping. To ensure data sovereignty, this study introduces a data synthesis methodology to bootstrap NMT models without scraping target-language parallel text. Focusing on Q'eqchi' Mayan, we transformed community-sourced dictionaries into a massive synthetic corpus, utilizing Parameter-Efficient Fine-Tuning (PEFT) via LoRA adapters on an mT5-base model. In-domain evaluation demonstrates high structural acquisition (BLEU 42.02), proving that synthetic constraints effectively teach complex agglutinative morphology and VOS word order. However, evaluation against an organic glossary reveals a structural-semantic gap (BLEU 0.59), where the model maintains grammatical integrity but lacks the lexical grounding of natural language. The model exhibits overfitting to the constrained structural variance of the synthetic templates; despite high semantic entropy in the pipeline, it struggles with the syntactic fluidity of natural language, forcing organic inputs into rigid learned patterns. Furthermore, an ablation study utilizing a Multi-Task Learning architecture resulted in negative transfer, suggesting that auxiliary tasks competed for limited parameter capacity within the LoRA adapters, causing over-optimization for synthetic markers at the expense of organic flexibility. Ultimately, we establish that synthetic bootstrapping is a highly effective structural primer, but requires authentic data for semantic refinement via Curriculum Learning.

Parametric imitation learning via behavior cloning can suffer from poor generalization to out-of-distribution states due to compounding errors during deployment. We show that reusing the training data during inference via a semi-parametric retrieval-based imitation learning approach can alleviate this challenge. We present Difference-Aware Retrieval Policies for Imitation Learning (DARP), a semi-parametric retrieval-based imitation learning approach that addresses this limitation by reparameterizing the imitation learning problem in terms of local neighborhood structure rather than direct state-to-action mappings. Instead of learning a global policy, DARP trains a model to predict actions based on $k$-nearest neighbors from expert demonstrations, their corresponding actions, and the relative distance vectors between neighbor states and query states. DARP requires no additional assumptions beyond those made for standard behavior cloning -- it does not require additional data collection, online expert feedback, or task-specific knowledge. We demonstrate consistent performance improvements of 15-46% over standard behavior cloning across diverse domains, including continuous control and robotic manipulation, and across different representations, including high-dimensional visual features. Code and demos are available at https://weirdlabuw.github.io/darp-site/.

Vision-Language-Action (VLA) models have demonstrated impressive end-to-end performance across a variety of robotic manipulation tasks. However, these policies offer no guarantees against collisions with task-irrelevant objects in the scene. Existing safety filters sidestep this problem by querying a vision-language model (VLM) to identify obstacles and their locations. This, however, is too slow to run in the control loop and can only be invoked at episode initialization, leaving the filter unable to track moving obstacles. We discover that a small number of attention heads within a VLA model reliably localize the object the policy intends to approach. These heads can be exploited within a training-free safety framework that obtains the active target from the attention heads at every step, treats the remainder of the scene as obstacles, and feeds these into a Control Barrier Function (CBF) filter. Together with a lightweight real-time object tracker, this allows for collision avoidance for non-static obstacles. We evaluate our framework on SafeLIBERO, which we extend with moving obstacles. On the original static benchmark, our method performs comparably to an oracle that uses privileged simulator state to identify the target, emulating a VLM-based identification step run once at episode initialization. On the dynamic variant, where the oracle's init-time target assignment becomes stale, our method substantially outperforms it by 43%, on average. Our findings suggest that the perceptual signals needed for real-time safety filtering are already present within VLA policies and can be exploited without additional training or heavy auxiliary models.

Vision-Language-Action (VLA) models demonstrate strong perfor-1 mance on language-conditioned robotic manipulation within their training dis-2 tribution, yet their generalization capabilities remain fundamentally limited. They3 lack the robustness required to handle perturbations, frequently failing when con-4 fronted with lighting changes, altered camera viewpoints, or small initial-state5 variations. We propose PROBEACT, a training-free runtime intervention frame-6 work that detects and recovers from grasping and placement failures in pre-7 trained VLA policies without modifying their weights or requiring additional8 demonstrations. PROBEACT combines three components: (i) a lightweight multi-9 target hidden-state probe that predicts the 3D positions of task-relevant objects10 from intermediate VLA features, with Hungarian-matched identity tracking for11 multi-object scenes; (ii) an object-agnostic kinematic state machine that detects12 grasp, transport, and placement failures using only gripper-internal signals and13 end-effector kinematics; and (iii) a hierarchical Control Barrier Function (CBF)14 filter that encodes repeated-failure locations as soft safe-set constraints, mini-15 mally correcting VLA actions while preserving baseline behavior. As a plug-and-16 play, training-free intervention loop, PROBEACT is orthogonal to existing train-17 ing pipelines. Evaluated on the LIBERO-plus benchmark, our framework acts as18 a universal safety net, improving the success rate of the OpenVLA-OFT model19 from 69.6% to 74.1%, while demonstrating broad applicability to both base and20 fine-tuned VLA policies.

Accurate dynamics models are essential for model-based robotic control, yet nominal Euler--Lagrange models often become inaccurate in the presence of payload variation, unmodeled coupling, friction, aerodynamic effects, and changing operating conditions. Most learning-based correction methods improve prediction accuracy by introducing a single additive residual, but do not preserve the internal mechanical structure of Euler--Lagrange systems. This leads to models that do not preserve symmetry, positive-definiteness, or the coupling between inertia and velocity-dependent terms, which can result in physically inconsistent predictions and reduced reliability when embedded in model-based controllers. We propose a structure-preserving residual learning framework that decomposes model mismatch into an inertia correction, the corresponding induced Coriolis term, and a generalized-force residual. The mechanical component is learned under physical constraints, while the disturbance-sensitive component is represented through a sparse history-dependent latent interaction model and adapted online using Bayesian linear regression. This separation preserves key mechanical structure while restricting adaptation to the part of the dynamics most affected by changing conditions. Experiments across multiple robotic platforms, including mobile, aerial, and manipulator systems, show that the proposed method improves dynamics prediction and trajectory tracking under coupled and time-varying dynamics. These results highlight the value of combining structured residual modeling, compact latent interaction selection, and selective online adaptation for real-world model-based control.

Vision-language-action (VLA) policies provide strong priors for language-conditioned manipulation, but remain brittle in off-nominal states requiring targeted recovery. We propose ReCoVLA -- a failure-conditioned residual recovery framework that keeps a pretrained VLA policy frozen, uses an external vision-language model (VLM) to infer the failure mode and recovery stage, and compiles a structured reward from task-relevant components. Rather than using the VLM to generate actions or rewards directly, ReCoVLA uses it as a semantic reward selector: it predicts a recovery descriptor and reward mask for in-simulation residual-policy training, followed by zero-shot sim-to-real deployment of the trained recovery policies. This decouples high-level failure understanding from low-level corrective control to support different VLAs. Experiments across short-horizon, long-horizon, and contact-rich manipulation tasks show that ReCoVLA outperforms the tested baselines on average. In simulation, our reward compiler improves average success from 36.7% for the fine-tuned $π_{0.5}$ baseline to 66.7%. In physical zero-shot sim-to-real experiments, ReCoVLA achieves the best average performance, with 61.7% success.

Dexterous manipulation presents substantial challenges for imitation learning due to its high-dimensional action space and complex contact-rich dynamics. Policies trained purely from demonstrations often suffer from compounding errors during deployment and require large amounts of expert data to achieve reliable performance. To move beyond the limitations of demonstration data, in this work, we propose DexPIE, a post-training framework for dexterous policy improvement from experience collected through real-world deployment. First, DexPIE enables effective exploration coverage through a dexterous-hand-adapted intervention system and multi-stage DAgger-style data collection across initial and intermediate task stages, providing reliable supervision for accurate policy evaluation. To reduce temporal noise between post-training rollouts and demonstration data, we introduce asynchronous inference in the relative action space, which better aligns rollout data with demonstrated behavior and allows the critic to learn a value function induced by a more consistent underlying policy. Finally, DexPIE improves the policy through conditioning on a continuous optimality indicator, allowing the policy to leverage the quality of data in a more fine-grained manner. Across three challenging real-world dexterous manipulation tasks, DexPIE achieves a 37% improvement in success rate over the demonstration-based reference policy, outperforming all baseline methods and demonstrating stronger robustness. The source code and dataset will be made publicly available.

Cooperative object transportation is essential in numerous domains, including industrial to domestic services. A popular transportation strategy is to carry objects on top of multi-robot systems. The corresponding task is typically solved by decomposing it into three interconnected subproblems: formation control, cooperative navigation, and collision avoidance. A particular challenge posed by real-world objects is their potentially arbitrary shape and non-uniform mass distribution, necessitating robot formations that securely support the object. In this work, we address the challenge of pattern formation control for transporting such real-world objects by proposing a novel multi-agent reinforcement learning approach. Our approach enables a multi-robot system to autonomously position itself underneath an object to support its weight while avoiding obstacles during the formation process. Our evaluations with diverse environments and varying numbers of robots show that our approach leads to policies that reliably produce balanced formations and generalize to cluttered scenes and objects with complex geometry and non-uniform mass distribution.

Vision-language-action models have shown strong promise for robot manipulation, yet raw language is primarily needed to specify task intent rather than to be repeatedly processed during high-frequency low-level execution. Motivated by this separation, we propose a cerebello-thalamic-inspired vision-action model (CT-VAM) for efficient task-conditioned visuomotor control. CT-VAM acts as a compact local execution policy that predicts action chunks from dualview visual observations, proprioception, and a lightweight task condition, potentially enabling a practical cloud-edge paradigm in which high-level semantic reasoning can be handled by large models while fast closed-loop control runs on local hardware. To fuse heterogeneous inputs effectively, CT-VAM introduces TARS (Thalamic Action Routing Stream), a stream-separated conditional attention decoder that independently routes action, visual and task streams, preventing dense sensory tokens from overwhelming compact task-relevant conditions. With only 68M parameters, CT-VAM achieves LIBERO success rates competitive with substantially larger VLA models, while reducing inference latency. Together with flow-consistent inpainting for asynchronous chunk execution, CT-VAM supports high-frequency control and demonstrates robust realworld deployment on resource-constrained robotic platforms.

With the advancement of visual sensing systems, computer vision is playing an increasingly important role in autonomous driving and robot navigation. Relative pose estimation in multi-camera systems is essential for accurate vehicle localization and environment perception, demanding high real-time performance and robustness. Existing methods, however, often involve high computational costs and rely heavily on abundant feature matches, limiting their applicability in time-sensitive driving scenarios. To address these limitations, this paper introduces a unified framework for efficient relative pose estimation, built upon a novel translation parameterization and first-order rotation approximation. Within this framework, we propose three efficient minimal solvers specifically designed for autonomous vehicles. The first solver integrates the vertical direction prior from Inertial Measurement Units (IMUs), the second utilizes the rotation axis direction prior during steering maneuvers, and the third is designed for planar motion - a realistic assumption for ground vehicles operating on structured roads. By reducing both the minimal number of point correspondences and the algebraic complexity, our methods enable faster hypothesis generation within RANSAC-based pipelines, improving suitability for real-time systems. Extensive experiments on synthetic datasets and the KITTI autonomous driving benchmark demonstrate that the proposed solvers achieve a favorable balance between speed and accuracy compared to existing state-of-the-art algorithms.

Offline safe reinforcement learning (Safe RL) enables policy learning without online interactions, making it suitable for safety-critical systems such as robotics systems. However, its reliance on static datasets exposes offline Safe RL to data poisoning attacks, where adversaries inject malicious samples that compromise safety and induce unsafe policy behavior. In this work, we propose a new learning paradigm, named safe reinforcement unlearning (Safe-RULE), used as a defense framework to remove the influence of poisoned data without retraining from scratch or requiring access to the original training environment. We further extend reinforcement unlearning to offline Safe RL by explicitly accounting for both task performance and safety constraints during the unlearning process. Experiments across benchmark Safe RL tasks demonstrate that our approach effectively enhances safety performance against data poisoning attacks.

Optical trapping and micromanipulation rely on carefully shaped electromagnetic fields to exert precise forces and torques on microscopic particles. Despite their widespread application in biology and nanotechnology, the absolute physical limits of trapping performance, specifically the maximum achievable optical force and trap stiffness, have not yet been rigorously quantified. This work establishes a general theoretical framework to determine these fundamental bounds in the dipole approximation. By relating the optical force and stiffness to a local Taylor expansion of the electromagnetic field at the particle location, we formulate the performance limit as a solution to a quadratically constrained quadratic program. To evaluate these bounds, we employ two complementary approaches. First, we utilize a complete basis of vector spherical wave functions to determine the absolute theoretical limits of optical force and stiffness permitted by Maxwell's equations in free space, revealing Pareto-optimal trade-offs between stable confinement and directional force. Second, we introduce an aperture-based formulation that restricts the incident fields to those realizable by finite planar apertures. This yields device-consistent bounds directly applicable to experimental setups which rely mostly on electromagnetic beams. The finding that optimized aperture fields can outperform standard Gaussian beams by removing the severe axial bottleneck is particularly important. By comparing these two regimes, we identify the specific spatial modes that contribute to stable trapping and quantify the performance trade-offs inherent to physical beam shaping. This dual framework provides provably optimal bounds for power-normalized optical tweezers and serves as a rigorous benchmark for evaluating realistic beam designs.

We study a Mach-Zehnder interferometer (MZI)-based optoelectronic magnonic parametric oscillator (OEMPO) incorporating a YIG-loaded magnonic branch and a tunable phase-shifter branch, enabling systematic investigation of adaptive interferometric oscillator dynamics under distributed phase perturbations. Through analysis of nondegenerate OEPO mode pairs and frequency-pulling behavior, the loop free spectral range (FSR) and effective delay time were quantitatively extracted. Despite the nominally frequency-pinned parametric operation, weak frequency pulling and OEPO mode softening were observed, revealing an additional adaptive interferometric degree of freedom introduced by the MZI architecture. By comparing local and global sampling configurations, we demonstrate that the YIG branch behaves predominantly as a local dispersive resonant subsystem governed by the complex magnonic susceptibility, whereas the phase-shifter branch primarily controls the global interferometric redistribution geometry. Nevertheless, coherent recombination and adaptive regeneration within the loop produce finite cross-coupling between the two branches, resulting in partially synchronized interferometric dynamics and branch-dependent adaptive redistribution. Quantitative complex-Lorentzian analysis further reveals substantial phase-to-amplitude conversion and distinct differences between the OEO and OEPO regimes: the phase-pinned OEPO favors strongly dispersive local YIG response, while the frequency-adaptive OEO exhibits more mixed absorptive--dispersive behavior due to spectral relaxation through frequency pulling. Broadly, the present platform establishes a versatile framework for exploring adaptive nonlinear interferometric physics, coherent phase redistribution, and branch-dependent synchronization phenomena in hybrid magnonic-photonic oscillator systems.

World models have recently seen a rapid growth in both their popularity and capability as more data efficient tools for generating robot training data or simulating real world environments, with many works proposing their integration into the robot learning pipeline. While highly practical, in this work we demonstrate that world models introduce a uniquely stealthy and effective data poisoning entry point into the robot learning supply chain that can result in the deployment of unsafe or otherwise compromised robotic policies despite training on seemingly safe ground truth training data. In contrast to traditional data poisoning techniques which directly implant dangerous trajectories into sold or uploaded datasets, our novel attack methods inject malicious prompts or compromising transition dynamics into visibly safe teleoperated datasets which are only activated once fed through a world model as input. This can result in the generation of synthetic, dangerous robot training trajectories and subsequently unsafe or compromised robot policies. We demonstrate the effectiveness of our attacks against both state of the art action conditioned and text conditioned world models, showing a full end-to-end backdoor on a downstream DRL policy and a proof-of-concept for the VLA setting. Overall these findings necessitate research into more secure world models and reevaluating their position within the robot learning supply chain.

Embodied policies typically map current observations directly to actions, leaving candidate-action consequences implicit. World models provide predictive supervision, representations, or external simulation, but rarely let a policy inspect the imagined consequence of its own proposal before acting. We introduce $ω$-EVA, a latent interactive world model that realizes an Envision--Verify--Act loop for embodied action generation. Its three-stage framework learns action-conditioned latent dynamics, trains a language-conditioned flow policy on dynamics-aware visual representations, and feeds the policy's proposal back through the world model. A tri-branch refiner jointly reasons over the current state, proposal-conditioned future, and proposed action to produce the final action chunk. Because consequence reasoning remains in latent feature space, $ω$-EVA avoids generating future videos at inference. Evaluations across diverse single-arm, bimanual, long-horizon, and perturbed simulation settings show that the complete interaction pipeline consistently improves the proposal policy, while latent diagnostics indicate meaningful action-conditioned future structure. With approximately 1.2B parameters and no additional robot-data pretraining, $ω$-EVA demonstrates a compact and competitive performance--scale--data trade-off, making the world model an active action-feedback module rather than a passive predictor.

Humans rely on spatially dense, geometry and force-aware tactile feedback at high temporal resolution for dexterous manipulation. While vision-based tactile sensors enable dense force estimation, they are limited by camera frame rates, motion blur, and data bandwidth. Event-based optical tactile sensors offer an attractive alternative with microsecond temporal resolution and low motion blur, but existing methods are restricted to predicting only net forces. We introduce the first framework for dense 3D force field reconstruction using event-based optical tactile sensors. Our approach estimates 3D surface displacements from event data and maps them to forces via the inverse Finite Elements Method (iFEM). Shear displacements are recovered through the proposed event-based marker tracking algorithm, while normal displacements are predicted by a convolutional neural network trained on a collected dataset of synchronized force-displacement-event data. Experiments demonstrate accurate reconstruction of physically grounded forces, achieving a mean absolute error of (0.14 N, 0.10 N, 0.93 N) over force ranges up to (4 N, 4 N, 20 N), while operating at an average of 100 Hz. This work constitutes a first step toward enabling dense force feedback for high-frequency control in robotic grasping and dexterous manipulation.

Robot middleware faces a new role in the era of Physical AI. Learned policies, planners, and vision-language-action (VLA) models now enter deployed robots as causal participants on the control path, but the layer that integrates them with timing, scheduling, and network has not been named. Recent language-agent work names this layer the harness, the external system that mediates tools, manages state, bounds resources, and records execution. The robotics community has not yet adopted this framing, and we propose that robot middleware is that harness. A Physical AI harness differs from a software harness in where it intervenes. A software harness mediates at tool-call boundaries. A Physical AI harness must mediate at control, computing, and communication simultaneously, because a learned policy's output crosses all three: its commands shift the trajectory, its inference time shifts the schedule, and its payload shifts the bandwidth. Robot middleware is the lowest robot-stack layer with mediating abstractions over all three, so it is best positioned to compose their enforcement. It already provides most of what a harness needs but lacks the enforcement for an AI model. We name this missing enforcement as three functions: Projection gates each output at emission, Isolation bounds the model's execution and transmission slot, and Transfer falls back to a verified baseline when checks fail. Each appears today as hand-built application code in deployed robot systems, built on surfaces robot middleware already provides. Robot middleware should host them not as the best single-axis enforcer but as the layer that composes all three. We sketch this as a ROS 2 Harness Profile, a deployment artifact that carries an AI model's declared output region, inference budget, and operating regime while the middleware enforces them across ROS 2, DDS, and Zenoh.

Body movement communicates intent at distances and in conditions where neither the face, nor speech can be captured. We study the recognition of communicative intent from 2D body pose alone. We argue that body motion is a reliable signal especially in scenarios that require real time low-cost on-device person-to-robot communication in long distance environments, such as rescue missions. However, existing resources do not isolate this signal. Affective corpora combine body, face, voice and text, while skeleton action-recognition benchmarks label the action performed rather than the message conveyed. We release a dataset of real frames of full-body pose covering ten communicative intents and we compare it against other real (IPC) and synthetic (MotionLCM, VEO3.1, Kimodo) ones that span a range of difficulty. We target systems that can run on a robot's limited onboard hardware. We benchmark multiple models, from skeleton graph classifiers to joint motion-forecasting networks, and report performance metrics together with frame rate on an embedded GPU (NVIDIA Orin~Nano), since speed matters as much as accuracy in our scenario. Finally, we show that a model's own autoregressive self-consistency works as an unsupervised reliability signal. We give a short proof that bounds the probability that a self-consistent prediction is correct, show that this probability grows with the number of consistent steps, and identify the conditions under which a confident prediction can still be false, benchmarked against industry-standard metrics.

In mixed-traffic environments where autonomous and human-driven vehicles may co-exist, motion planning for autonomous vehicles requires anticipating the future behaviors of surrounding human drivers. Existing reinforcement learning-based methods generally directly incorporate the predicted human intents into the observation to enable a proactive planning. However, human intent is inherently uncertain due to the behavioral diversity, perception noise, and partial observability. Treating predicted intends as deterministic states can result in unsafe decisions for autonomous vehicles. To address this problem, we propose Uncertainty-Aware Motion Planning (UAMP), which incorporates uncertainty in human intent prediction for AV decision-making. Specifically, UAMP first introduces a proximity-aware uncertainty estimator to quantify the interaction-conditioned intent uncertainty and constructs an uncertainty-guided joint intent distribution over surrounding human-driven vehicles. Within this uncertainty set, UAMP further introduces Uncertainty-Calibrated Value Learning (UCVL) to correct value function learning biases arising from directly incorporating uncertain human intent predictions into the observation. Extensive experiments in various mixed-traffic scenarios show that UAMP significantly improves safety and driving comfort, while maintaining traffic efficiency compared with existing approaches. The code is released at https://anonymous.4open.science/r/UAMP-5638.

Learning robot manipulation policies with deep neural networks from a single demonstration remains highly challenging, as even small deviations from the demonstrated trajectory can quickly compound into failure, while collecting substantial online interaction data is costly. We propose ReGIL, a retrieval-guided imitation learning framework that treats a single demonstration as an external memory. ReGIL repeatedly queries this static memory throughout training to simultaneously guide exploration, generate the regularization buffer, and construct rewards. Specifically, it computes rewards through local temporal alignment between the current trajectory and the retrieved segment, providing step-wise and informative feedback for policy improvement. We evaluate ReGIL on robotic manipulation tasks from the LIBERO and Meta-World benchmarks under the single demonstration setting. ReGIL outperforms prior baselines in both success rate and training efficiency. In real-robot experiments, using only one demonstration and less than one hour of online training, ReGIL achieves over 75% success rate across three manipulation tasks with randomness in both initial robot pose and target position. These results demonstrate that leveraging the single demonstration as reusable memory can provide more than static supervision for efficient robot learning. More details can be found on our website: https://regil2026.github.io/

Vision-Language-Action (VLA) models have become a powerful framework for robotic manipulation, and recent studies have introduced tactile or force feedback into VLAs to address contact-rich tasks. However, these models are typically deployed as offline policies. When contact conditions shift from the training distribution, the policy cannot perform online adaptation, leading to problems such as inappropriate contact forces and inefficient retries. Therefore, we propose TORL-VLA, a tactile-guided online reinforcement learning framework that couples tactile feedback with policy refinement for contact-rich manipulation. Our method introduces a tactile-derived wrench-aware VLA to predict reference actions and future wrench sequences, while a lightweight online RL module is used to refine the reference actions. To stabilize learning from mixed exploratory policy-generated and human-intervention data, we introduce an intervention-censored critic that prevents post-intervention success from being wrongly credited to policy-generated actions preceding intervention. Real-robot experiments on long-horizon contact-rich tasks, including latch manipulation, coffee-cup placement, and egg handling, show that TORL-VLA improves success rates at both subtask and full-task levels, as well as time-bounded execution efficiency over strong baselines.

Generating high-quality dexterous grasps remains challenging for learning-based methods, which often depend on carefully tuned contact losses or costly contact-based test-time refinement. We present KPGrasp, a flow-matching framework that learns dexterous grasp priors from large-scale data rather than relying on contact losses or contact-based test-time refinement. KPGrasp couples an all-Euclidean 3D hand-keypoint parameterization with a simple yet scalable Transformer flow model. The parameterization avoids the drawbacks of the conventional mixed SE(3) pose and joint-angle output space, expresses grasps in the same frame as the object point cloud, and thus enables native spatial reasoning; the Transformer flow model is trained with only the standard flow-matching loss and scales effectively with data, model capacity, and batch size. Experiments demonstrate state-of-the-art performance on two simulation benchmarks. On the Dexonomy benchmark, it reaches a 76.3% grasp success rate, improving over the strongest directly comparable baseline by 47.4% while reducing penetration depth to 2.4 mm. The same model also achieves the best average performance on the DexGrasp Anything benchmark without fine-tuning. For batched inference, KPGrasp requires only 0.032 s per grasp. Finally, real-world experiments on 20 diverse objects demonstrate that the pipeline can be deployed in a real-world setup.

Reliable navigation in GPS-denied environments remains a fundamental challenge in robotics, aerospace, and autonomous vehicle applications. This paper presents a Dual Quaternion-Based Unscented Kalman Filter (DQUKF) equipped with a Visual Inertial Odometry (VIO) algorithm for accurate state estimation enabling navigation in GPS denied locations. The proposed framework formulates the DQUKF in an error state manner, where the nominal pose is represented by a unit dual quaternion and the local pose error is represented by a 6-dimensional twistor parameterization used for sigma point generation, covariance propagation, and measurement correction. In parallel, the VIO algorithm tracks features across image frames, synchronizes measurements between the IMU and camera, and provides visual constraints that complement inertial propagation. Simulation results on the EuRoC MAV dataset show that the proposed DQUKF converges under high initialization uncertainty and achieves a position RMSE of 0.2584~m in the difficult flight sequence, outperforming the benchmark filters.

Humanoid robots hold immense potential for real-world assistance, yet agile interaction with objects in unstructured environments demands tightly coupled whole-body coordination. Despite recent advancements, current controllers face a critical deployment gap. They rely heavily on dense reference trajectories and perfect state observability, which inherently limits physical generalization. We present Vision Guided Agile Interaction Control (VAIC), a unified framework that bridges this gap by operating exclusively on onboard depth, historical proprioception, and a decoupled user command interface. VAIC employs a two-stage distillation paradigm. First, a privileged teacher policy masters diverse interaction skills using precise object kinematics and exact environmental states. Second, a deployable student policy distills these capabilities by replacing full body tracking with velocity targets across multiple axes and an interaction indicator for each frame. The student utilizes a recurrent object adaptation module to implicitly infer unobservable object dynamics from raw depth streams and proprioception. Evaluations and real-world deployments on the humanoid robot demonstrate that a single VAIC policy successfully executes highly diverse dynamic tasks. These tasks include box carrying, cart interaction, and skateboarding, consistently outperforming baselines and advancing autonomous humanoid deployment.

Vision-language-action (VLA) policies can deviate from nominal trajectories during manipulation, even when tasks remain physically feasible. Recovering from these deviations is challenging, as they push the policy into unfamiliar state spaces where direct re-planning frequently destabilizes action sequences. We propose Back to the Familiar Future (B2FF), a recovery framework for foresight-driven VLAs that leverages future visual conditioning as a recovery interface. Before execution, the VLA generates a milestone bank of familiar future states conditioned on the clean initial observation. At recovery time, a recoverability-aware selector selects a recovery milestone from this bank and enforces it as a fixed visual goal. This enables the VLA to robustly map off-trajectory observations back to a familiar future. On failure-injected LIBERO, under controlled recovery timing aligned with the injected failure, B2FF increases the average success rate of a baseline VLA from 56.3% to 74.0%, demonstrating that pre-imagined milestones can guide recovery without fine-tuning the low-level action generator.

We present RPO-PDT: a retrieval-grounded, role-play-based dialogue system for adaptive student support in higher education. RPO-PDT is: (1) able to provide institution-specific Personal Development Tutor (PDT) guidance using structured knowledge sources; (2) constrained by explicit persona, boundary, confidentiality, and safety policies; and (3) designed around a reverse-roleplay loop where unresolved interactions are replayed from the student perspective, enabling alternative tutor strategies to be generated and stored as reusable strategy memory. RPO-PDT supports both text-based and Furhat-based embodied interaction for demonstrating grounded, safe, and adaptive student-support dialogue.

Time-of-flight cameras are popular in robotics for providing direct depth information while being compact, inexpensive, and robust to lighting conditions, but their low spatial resolution and depth noise are widely believed to preclude precise feedback control. In this paper, we show that an inexpensive, low-resolution time-of-flight camera provides sufficient feedback to reliably and precisely balance an inverted pendulum on a cart--a canonical benchmark for fast, unstable dynamics.

Autonomous Racing has seen remarkable progress through deep Reinforcement Learning (RL), primarily for four-wheeled vehicles. However, motorbikes introduce substantially greater complexity due to the need to manage balance and lean angle, in addition to more reactive steering and throttle control, and a smaller weight. In this work, we present a framework for training an autonomous agent to race a superbike in VRider SBK, a physics-accurate Unity-based motorbike simulator. Our approach integrates Soft Actor-Critic (SAC) with Self-Paced curriculum Deep reinforcement Learning (SPDL), which dynamically generates progressively more challenging tasks based on the agent's performance, without requiring manual curriculum design. The agent's state space comprises proprioceptive features extended with lean-angle history, along with global track features via course points. The reward signal is shaped to encourage progress along the track while penalizing instability-inducing behaviors specific to two-wheeled dynamics. Preliminary experimental results demonstrate that SPDL outperforms SAC alone in training efficiency, lap time, and driving stability across multiple tracks and motorbike models, establishing a first baseline for RL-based autonomous motorbike racing.

World Action Models (WAMs) couple a video dynamics prior to the policy and have shown encouraging results on tabletop manipulation, but iterative denoising over high-dimensional video-action latents leaves them too slow for real-time humanoid loco-manipulation. The problem is compounded by the dominant hierarchical paradigm, in which a high-level manipulation policy controls only the upper body while a low-level controller tracks coarse base commands -- placing upper and lower body in inconsistent action spaces and reducing the legs to balance-preserving locomotion. We present MotionWAM, a real-time WAM that drives autonomous humanoid loco-manipulation from a single egocentric camera by conditioning the policy on the intermediate denoising features of a video world model. MotionWAM replaces the upper-lower split with a unified motion latent and predicts whole-body motion tokens that jointly cover locomotion, torso motion, height regulation, foot interaction, and hand manipulation in a single action space. A three-stage learning framework progressively adapts the video world model to egocentric visual dynamics and to the target humanoid embodiment. On nine real-world Unitree G1 tasks, MotionWAM runs in real time, substantially outperforms Vision-Language-Action (VLA) baselines fine-tuned on the same demonstrations by over 30% in overall success rate, and executes task-driven foot interaction that decoupled upper-lower policies cannot reach. Our results suggest that video-pretrained WAMs can be lifted from tabletop manipulation to coordinated, human-like whole-body humanoid control.

Autonomous obstacle removal from the ground is an important earthwork task, but this is difficult to automate because an excavator must adapt its excavation trajectories over repeated cycles as soil-obstacle conditions change. Learning such state-dependent behavior requires a training environment that reproduces accumulated soil-obstacle interactions, including contact states, terrain deformation, and obstacle visibility. Accordingly, particle-based simulation is suitable for the relevant policy learning. However, particle simulation is computationally expensive, and repeated excavation cycles further increase the learning cost. We observe that the burial condition of an obstacle governs both task difficulty and simulation cost: deeper burial makes obstacle removal harder while also requiring more particles for accurate simulation. This observation motivates a burial-conditioned curriculum learning strategy. We propose a time-efficient sim-to-real policy learning framework in which the policy observes terrain and obstacle information from RGB-D measurements and then outputs a parameterized excavation trajectory; in this process, the simulator reproduces in a real-world excavator the same observation-action interface it uses under controllable burial conditions. The curriculum begins with shallow burial conditions and progressively increases burial depth while adjusting particle count, thus simultaneously controlling task difficulty and simulation cost. Experiments show that the proposed framework successfully learns an effective obstacle-removal policy, whereas baseline methods fail even after a full week of training. The proposed curriculum achieves effective performance within three days and achieves successful transfer to a real 12-ton excavator operating on open ground with various steel obstacles, thus demonstrating robust obstacle removal.

Social Learning is a decentralized decision-making paradigm in which spatially dispersed agents collect streaming observations regulated by one of a finite number of models (the hypotheses). The agents are interested in assigning probability scores (the beliefs) to the possible hypotheses. To this end, the agents exchange their beliefs according to a certain communication graph. It has been shown that, under reasonable conditions on the identifiability of the decision model and the network connectivity, each agent ultimately places all the belief mass on the true hypothesis governing the data. However, several questions remain unanswered regarding the evaluation of the social learning performance. One recently adopted performance metric is the rejection rate, i.e., the rate at which the beliefs about the erroneous hypotheses vanish. One contribution of this work is to establish that the rejection rate leads to several paradoxes, which make it unsuitable as a valid performance measure. We then focus on studying the error probability measure. For a binary Gaussian problem, we derive an analytical formula characterizing the ratio between the individual agents' probabilities and the optimal Bayesian probability. The formula shows that this ratio is expressed by the product of two terms quantifying the effect of the network connectivity and the role of the prior information. As a result, an irreducible gap emerges between the decentralized and the centralized error probabilities, which is agent-dependent and does not disappear asymptotically.

Cloud network infrastructure at hyperscale presents unique operational challenges where traditional human-driven incident response cannot keep pace with the volume, velocity, and complexity of failures. This paper presents an agentic AI architecture for autonomous incident resolution in large-scale network operations. Our system employs a multi-agent orchestration framework where specialized AI agents collaborate to detect, diagnose, and remediate network incidents without human intervention. We describe the architectural principles, including hierarchical agent decomposition, skills-based tool invocation via standardized protocols, structured knowledge encoding from operational runbooks, progressive autonomy with safety boundaries, and closed-loop verification. The architecture has been deployed in production at a major cloud provider, demonstrating that agentic AI systems can achieve autonomous resolution rates exceeding 90% for common incident categories while maintaining safety guarantees through layered authorization and rollback mechanisms. We discuss design tradeoffs, failure modes, and lessons learned from operating autonomous AI agents at scale.

Plasmon-exciton hybrids, or plexcitons, offer deeply subwavelength light-matter interactions with versatile pathways for energy redistribution. Incorporating chirality into such systems is particularly compelling, enabling spin-sensitive optical functionality that can operate on ultrafast timescales and within ultracompact volumes. Despite recent progress in chiral plexcitonic systems, how structural chirality and plasmon-exciton coupling determine chiroptical spectra and ultrafast energy flow remains elusive. Here we realize chiral plexcitons by functionalizing intrinsically chiral gold helicoid nanoparticles with molecular J-aggregates. Within a non-Hermitian framework, we trace the microscopic origin of the helicoid chiroptical response and its coupling to the excitonic transition, revealing how the helicity of light selectively addresses distinct hybrid responses. At the spatiotemporal extreme, we find that the gap-localized response not only enhances polarization-sensitive contrast but also strengthens the local hybrid interaction, leading to accelerated ultrafast relaxation. Together, these space-, time-, and polarization-resolved measurements provide a physically grounded and experimentally benchmarked picture of chiral plexcitonic coupling, identifying chirality as a practical control parameter for selectively steering nanoscale energy pathways and dynamics.

Interior permanent magnet synchronous motor (IPMSM) design requires balancing conflicting objectives and multi-physics constraints, while modern optimization workflows face three bottlenecks: manual problem setup, high finite element analysis (FEA) cost, and unreliable surrogate-based search in sparse or out-of-distribution regions. To address these limitations, we propose an end-to-end automated IPMSM design optimization framework that integrates retrieval-augmented generation (RAG) for structured problem definition with an uncertainty-aware FEA-AI hybrid optimization pipeline. A Design agent, connected to a motor textbook through RAG, provides domain-knowledge-based options and engineering tips, and compiles an optimization card and a design-of-experiments plan for AI-model training. A Training agent automates electromagnetic FEA, records geometry-validation and solver-failure logs, analyzes failed geometries using ANOVA-based data analysis and LLM reasoning, and invokes a Design Sampling agent to redefine the design space and generate additional samples. An Optimization agent performs GA-based search with uncertainty-driven switching: low-uncertainty candidates are evaluated by AI-surrogate inference, whereas high-uncertainty and reliability-critical Pareto-front or top-K candidates are corrected by high-fidelity FEA and reused for iterative retraining. The framework converts manual, experience-dependent configuration into a reproducible workflow that balances computational cost and prediction reliability. Experimental results under a matched high-fidelity FEA budget show that the proposed hybrid approach achieves better objective performance while maintaining low and further reducible predictive uncertainty, outperforming FEA-only search, which is limited by early budget exhaustion, and AI-only search, which converges to a low-confidence optimum.

Warm starts are widely used to accelerate AC optimal power flow (ACOPF) solves, but the impact of different initialization strategies has received limited systematic study, particularly for the primal-dual interior-point methods that dominate large-scale ACOPF algorithms. This paper benchmarks initialization strategies for ACOPF solved with the interior-point solver IPOPT on 19 PGLib-OPF instances (5 to 30,000 buses), testing all 15 non-empty subsets of the primal blocks $\{P_g, Q_g, V_m, V_a\}$ under oracle conditions and three DC-seeded combinations in a practical setting. The experiments show that most partial primal-plus-dual restarts increase solve time or reduce convergence reliability. Among the oracle primal-plus-dual (O-PD) configurations, only the complete restart reliably converges on every baseline-convergent case, reaching a $47.6\%$ median solve-time speedup. Twelve of the 14 partial O-PD combinations have negative median speedups, and several fail repeatedly on larger networks. Decomposing the dual into constraint and bound multipliers shows that \emph{coverage}, not the presence of duals per se, governs robustness: the full bound-multiplier vector reaches 90.7\% convergence and a $+26.8$\% median speedup, whereas block-matched coverage (oracle multipliers on some bounds, defaults on the rest) drops to 70.4\% and $-31.1$\%. Practical DC seeding sometimes helps the AC solve, but the benefit is no longer statistically significant once the DCOPF presolve cost is included in the end-to-end comparison ($p = 0.4171$). For learned warm-start methods, the results support the following target ordering: predict the full primal vector first; if only partial coverage is possible, prioritize voltage variables; and avoid partial or inconsistent dual predictions unless the primal estimate is nearly complete.

Agent benchmarks score submissions with outcome verifiers that are typically hand-written and brittle, leaving them open to reward hacking. We audit 1,968 tasks across five terminal-agent benchmarks and find 323 (16%) hackable by frontier models given only the task description. This corrupts both leaderboard rankings and RL training signal, yet the standard response is manual and reactive. We introduce the hacker-fixer loop, a method for building exploit-resistant verifiers without per-task manual patching. The loop alternates three LLM agents: a hacker tries to pass the verifier without solving the task, a fixer patches the verifier to reject each discovered exploit, and a solver confirms the patched verifier still admits legitimate solutions. The loop iterates: each patch reshapes what the verifier rewards, surfacing the next exploit. We further add verifier access, and let patches transfer across tasks, to broaden the exploits the loop discovers. On KernelBench, the loop drives the attack success rate from 62% to 0% on a held-out corpus of publicly reported exploits. We also find that weaker agents in the loop can defend against much stronger hackers: Gemini 3 Flash's loop drives the stronger Gemini 3.1 Pro and Claude Opus 4.7's attack success rate from 76% and 61% to 0% on KernelBench, and Gemini 3.1 Pro's from 39% to 17% on Terminal Bench across 77 tasks. We release Terminal Wrench (323 hackable environments, 3,632 hack trajectories) as a snapshot of the current attack surface, our patched verifiers, the exploits the loop discovered, and our implementation as a basis for future work.

This paper proposes a design-for-test (DFT) methodology and architecture for testing and validation of silicon photonic integrated circuits. We describe the design of silicon photonic circuits and components that comprise the proposed DFT architecture. The designs are extensively simulated and validated as test-access and fault-detection circuitry. We demonstrate how the DFT approach can be deployed on photonic integrated circuits and how they can be tested for correct operation, in terms of signal power and phase. The application is demonstrated on two distinct types of designs -- an optical neural network comprising optical devices in a feed-forward topology, and on an optical logic circuit with feedback loops.

Real-world LLM applications are moving beyond single-agent workflows toward orchestrated multi-agent systems, yet current models still struggle to determine what each sub-agent needs to know. To measure this, we introduce PerspectiveGap, a benchmark for evaluating LLMs' ability to compose orchestration prompts for multi-agent systems. PerspectiveGap contains 110 scenarios, each evaluated through two distractor-mixed task formats: role-fragment assignment and free-form prompt writing. These scenarios are organized into 10 topologies, which are distilled from the authors' real-world engineering practice and framed by the Prompt Economy principle: building loop-centered orchestrations that maximize utility with minimal role and engineering overhead. In experiments with 27 commercial models from 10 companies, GPT-5.5 substantially outperforms all competitors, whereas Opus 4.7 shows a notable weakness in orchestration prompting despite its strong coding performance. Nevertheless, PerspectiveGap remains challenging: the evaluated models achieve an average combined pass rate of only 14.9\% (GPT-5.5 62.0\%) and an average overall leakage rate of 246.5\% (a per-scenario information leak-event count, not a proportion; GPT-5.5 49.1\%). These findings suggest that multi-agent orchestration prompting is a distinct and under-evaluated capability, and PerspectiveGap provides a foundation for measuring and improving it systematically.

This paper presents an adaptive model predictive control (MPC) framework for nonlinear urban air mobility (UAM) vehicles operating across the full flight envelope. The proposed approach leverages a linear parameter-varying (LPV) representation to update the predictive model online, enabling accurate capture of strongly nonlinear and time-varying dynamics associated with distributed electric propulsion (DEP) eVTOL aircraft. To systematically address the high-dimensional and coupled nature of MPC tuning, a multi-objective evolutionary optimization strategy based on NSGA-II is employed, incorporating proper normalization of states and control inputs to ensure balanced weighting and meaningful exploration of the design space. The resulting controller explicitly accounts for actuator constraints and enables reconfigurable control allocation for fault-tolerant operation. The framework is evaluated in nonlinear simulations using NASA's Generic Urban Air Mobility (GUAM) model and benchmarked against a robust servomechanism linear quadratic regulator (RSLQR). Results demonstrate that the proposed adaptive MPC achieves improved trajectory tracking and enhanced robustness under both nominal conditions and actuator degradation scenarios, including partial motor failure, while maintaining constraint satisfaction throughout all flight regimes.

Autonomous agents negotiate, purchase, deploy code, and move funds, but no neutral mechanism determines whether they met their delegated obligation, who is responsible when they did not, or which settlement action follows. This is the agentic clearing problem. Tool protocols (MCP), inter-agent communication (A2A), payment rails (x402), mandate and network agent protocols (AP2, Visa, Mastercard), and settlement-risk standards each assume that determination and none produce it. Clearing is the missing primitive. Payment is not clearing. Authorization is not clearing. LLM-as-judge evaluation is not clearing. Settlement-risk escrow is not clearing: it consumes clearing decisions. RAILS (Real-Time Agent Integrity & Ledger Settlement) is the integrity and clearing layer for agentic commerce, spanning a per-output reliability score, a published reliability record, and a clearing function that consumes them. The clearing protocol at its core closes that gap. Seven primitives (Obligation Object, Evidence Envelope, Verification Mesh, Clearing Decision, Settlement Instruction, Clearing Passport, Finality Rules), bound by a formal model of admissibility-graded verification, together yield a soundness property: no financially material settlement is supported by evidence below the obligation's admissibility floor. The property is falsifiable against the spec. We are not aware of a prior agent-commerce verification mechanism that states a property of this kind. The approaches nearest to it emit a pass, a delivery guarantee, a bare score, or an equilibrium. This paper specifies that clearing protocol.

In teleoperation, the human operator typically controls only the end-effector pose, which often leads to self-collisions of the manipulator and collisions with environmental obstacles, since joints and links are not controlled individually. A common strategy to mitigate this issue is to enhance the operator's input using optimal-control-based trajectory planning. As derivative-based solvers require differentiable constraints, existing approaches either approximate robots and obstacles with spheres, reducing geometric accuracy, or approximate derivatives, degrading convergence and increasing computation times. We address these limitations by adapting a recent formulation of differentiable collision-avoidance constraints, based on duality in convex optimization, to the teleoperation setting. The robot is approximated with capsules and the environment with polytopes. We compare the resulting trajectory planning method against state-of-the-art techniques in simulation with varying numbers of obstacles and evaluate it on a UR5e manipulator in a real-world teleoperation test. Results show that our approach achieves lower computation times while enabling more accurate obstacle modeling, leading to smoother and collision-free end-effector teleoperation.

Multirotors are widely used in applications ranging from surveillance to precision agriculture, yet conventional designs remain limited by their under-actuation. Tilt-rotor configurations overcome this limitation by enabling full actuation. This paper investigates neural-network-based control strategies for a fully actuated tilt-rotor system with four thrust-vectoring inputs. Our work is structured in two parts. First, we deliberately present a negative result by evaluating a direct input-output control approach. In this method, multilayer perceptrons (MLPs), long short-term memory (LSTM) networks, and transformer models are trained to map system states and their desired values directly to control signals. We show that this strategy fails to stabilize the system, highlighting the inherent difficulty of applying direct input-output learning to highly unstable plants. Second, as the main contribution, we propose a neural-network-enhanced sliding mode controller (SMC). The method decomposes the system dynamics into input-independent and input-dependent components, with the former learned from a small dataset using lightweight networks, thereby reducing real-time computational demands. Moreover, the proposed method can be trained using flight logs collected from low-performance controllers, and the resulting dynamic model learned from real-world data can be used in simulation. We further compare MLP- and LSTM-based implementations under model uncertainties and external disturbances, demonstrating the robustness and effectiveness of the proposed approach; in particular, the controller with the LSTM plant dynamics predictor achieves superior performance to its MLP-based counterpart while also exhibiting lower runtime.

Protecting an asset against threats is a challenging problem in an era of continuously evolving intelligent attacks. This requires cooperation between the asset and the defender to share information and jointly maneuver. To address this problem, this work proposes a cooperative guidance and control strategy for active asset protection against a maneuvering threat. This work develops a joint maneuver strategy where both the defender and the asset coordinate their time-varying speeds and courses to neutralize/capture the attacker. The control strategy is formulated around three coupled geometric and temporal objectives. The first objective is to set the line-of-sight rate between the asset and the attacker to zero, putting the attacker on a collision course and reducing their maneuvering. The second objective is to maintain the defender on the line-of-sight between the asset and the attacker. This ensures that the attacker faces the defender first before reaching the vicinity of the asset. Lastly, the defender is also guided to pursue the attacker based on the time-to-go estimates between the defender and the attacker. While keeping these objectives in mind, the control actions for the asset and the defender are jointly designed, fostering cooperation between the two. The stability of the proposed strategy is established using a Lyapunov-based approach. Numerical simulations performed show the effectiveness of the proposed cooperative strategy in ensuring the successful capture of a maneuvering threat.

We study data-driven real-time economic optimization of a multi-product chemical reactor when no reliable first-principles model is available beyond a steady-state energy balance. Instead of learning the economic objective directly as a black-box function, we use a composite formulation in which Gaussian process (GP) models predict physically meaningful outputs, including product concentrations and reactor temperature, while profit is computed analytically from these predictions together with raw-material, product, and utility prices. This preserves the structure of the economic objective, makes it parametric in changing prices without needing retraining, and allows candidate operating points to be checked against the available energy balance through a physics residual. The GPs also provide predictive uncertainty, which is exploited in a Bayesian optimization (BO) framework both for data-efficient exploration and for conservative enforcement of the reactor temperature constraint through an upper confidence bound. The acquisition function additionally penalizes large energy-balance mismatch obtained by substituting the GP-predicted outputs and candidate inputs into the available steady-state energy balance. The approach is demonstrated on a benchmark simulation of a non-isothermal multi-product reactor. Relative to a trust-region safe BO implementation, the proposed method achieves better simulated economic performance within the available iteration budget. Relative to a purely data-driven BO approach that does not use the available physics information, it avoids reactor temperature constraint violations.

I discuss some quantitative representations of Promise Theory for processes involving autonomous agents. Agent models are common in software systems, machine learning, and biology, for example, but may also apply to physics and other forms of engineering. I describe how Bayesian probability and information theoretic optimization, including Active Inference, may be incorporated with promise semantics -- as well as how Promise Theory supplements solutions, helping to avoid probability's pitfalls, which include non-local coordination, calibrating, and normalizing probabilistic computations. The role of boundary conditions in constraining allowed states and selecting decision thresholds is a form of promise, and agent alignment provides a scalable definition of intent. Autonomous agents may congeal into swarms with superagent characteristics by trying to minimize their information, despite uncertainty that works to maximize it. The use of Promise Theory involves some research challenges as well as stylistic preferences.

Multi-agent LLM systems for medical question answering often treat consensus as a reliability signal: if multiple agents agree on an answer, it is presumed trustworthy. However, answer-level consensus does not entail reasoning-level alignment. We introduce CARA (Cross-Agent Reasoning Alignment), a family of automated metrics that measure whether agents who agree on an answer also agree on the reasoning. Applying CARA to a standard debate system on two medical QA benchmarks, MedQA-USMLE and MedThink-Bench, we identify the consistency illusion: a failure mode where debate reduces detectable contradictions between agents while simultaneously decreasing the semantic similarity of their reasoning chains; agents appear to agree more but reason less consistently. To improve this misalignment, we propose the Grounded Debate Protocol (GDP), a prompt-level intervention that requires agents to commit to named medical facts and take explicit stances on other agents' claims. GDP produces large, consistent alignment improvements, with Cohen's d ranging from +1.43 to +1.99, across two datasets and two backbone models, without adding LLM calls or modifying system architecture. Our results motivate cross-agent reasoning alignment as a quantity to audit alongside accuracy in safety-critical domains.

Generating complete 3D scenes from a single image requires inferring globally consistent geometry, object relationships, and environmental context from inherently ambiguous visual evidence. Despite recent progress in joint layout-and-mesh generation, existing methods often rely on holistic or weakly decomposed pipelines that entangle many factors at once and demand extensive scene-level supervision, limiting their generalization to complex real-world environments. We propose a multi-agent orchestration framework that decomposes single-image 3D scene generation into three structured stages: scene initialization, environment construction, and multi-agent refinement. The initialization stage extracts image-derived object masks, builds object-level 3D representations, and predicts an initial spatial layout to form a coarse 3D scene. The environment-construction stage then leverages this initialization together with point-map geometry to build an environmental scaffold of supporting surfaces, room boundaries, materials, and illumination. Finally, in the refinement stage, a planner agent identifies structural and visual inconsistencies, applies simple corrections directly, and dispatches specialist agents for complex localized revisions that are reintegrated into the global scene. To provide reliable structural initialization while reducing reliance on scene-level annotations, we further introduce a geometry-aware layout predictor supervised by sparse geometric priors derived from point maps. Unlike fully supervised layout generators, the predictor can be trained from segmentation-level data and generalizes robustly to diverse real-world scenes. Extensive experiments on benchmark datasets show that our method consistently outperforms prior approaches in geometric accuracy, spatial consistency, and perceptual realism.

Adaptive beamforming is a cornerstone of array signal processing, yet its performance often collapses in the face of complex, rapidly changing interference. When interferers appear or move unpredictably, conventional estimators encounter a fundamental memory trade-off: short windows enable rapid tracking but suffer from high estimation variance, while long windows provide stable rejection but fail to adapt to shifts. This challenge is resolved by introducing the Universal Switching Beamformer (USB), which integrates competitive sequential prediction into the beamforming architecture. By employing a linear transition diagram, the USB implicitly maintains an exponentially large family of candidate covariance histories and dynamically re-weights them based on their cumulative output power. This mechanism allows the beamformer to automatically vary its effective memory length without explicit change detection or heuristic parameter tuning. A theoretical upper bound is proven on the regret relative to an omniscient oracle that selects the best piecewise-stationary covariance model in hindsight. Extensive simulations and experiments on the SwellEx-96 dataset demonstrate that the USB achieves the agility of short-window estimators and the precision of long-term integration, providing a principled solution for tracking highly non-stationary scenes.

Most evaluations of LLM agents look like exams: a discrete task, a clean environment, a score in minutes or hours. We argue that this approach is mismatched with the deployment conditions of autonomous systems, where the relevant timescale can be weeks to months, and where the dynamics that matter most, such as behavioral drift, governance in diverse environmental contexts, and cross-influence between agents from different model families, only emerge over time. We introduce Emergence World, a continuously running multi-agent simulation platform designed to make those dynamics measurable. The platform hosts populations of LLM-driven agents in a shared spatial world grounded in live external data (e.g. real-time weather, news APIs, internet access), equips each agent with 120+ specialized tools and three persistent memory systems, and lets them govern themselves through democratic mechanisms with consequential outcomes. The platform is model-agnostic at the reasoning layer and supports heterogeneous populations in which agents from different vendors share the same world. To illustrate the kinds of questions the platform makes tractable, we present a 15-day cross-vendor study with five parallel worlds powered by Claude Sonnet 4.6, Grok 4.1 Fast, Gemini 3 Flash, GPT-5-mini, and a mixed population. Identical roles and starting conditions produced radically different outcomes, ranging from stable deliberative governance to total population collapse. We release the prompts, log data and configurations to support further research on long-horizon multi-agent autonomy.

As language models are increasingly deployed as autonomous agents, they must coordinate with others over long horizons in open-ended interactive tasks. Yet existing evaluations rarely test these demands together, instead emphasising single-agent tasks, short interactions, or highly structured multi-agent settings. We introduce $alem$, a JAX-based benchmark for open-ended multi-agent coordination built on Craftax-like dynamics. Alem embeds procedurally generated coordination tasks, soft specialisation, communication, and controllable coordination difficulty into a long-horizon survival world with exploration, crafting, trading, and combat. We evaluate $13$ modern LLMs zero-shot within homogeneous teams, with trained MARL agents as reference points. Current LLM agents remain far from solving alem, averaging only ~6% normalised return, but their failures are not uniform. On the hardest coordination setting, zero-shot Gemini-3.1-Pro-High approaches MARL agents trained for one billion steps, while GPT-5.4-High achieves strong base-task reward but much lower coordination reward. This contrast shows that individual task competence does not imply coordination competence. Ablations show that communication is the largest contributor to coordination, while memory and reasoning help when used to maintain multi-step plans. Overall, our results identify coordination as a distinct bottleneck for frontier LLM agents, separate from single-agent capabilities. Alem makes this bottleneck measurable and provides a controlled testbed for developing agents that communicate, allocate roles, and execute shared plans. Code is available at https://github.com/alem-world/alem-env.

This paper benchmarks sequential feedback optimization (SFO) for wind farm power maximization using a medium-fidelity dynamic flow model. We compare SFO with two well-established approaches, adjoint-based economic model predictive control (AMPC) and extremum seeking control (ESC), under a common nine-turbine layout and identical operating constraints. The comparison focuses on steady-state power production and computational efficiency, both relevant for real-time implementation. The simulation results illustrate that SFO achieves higher steady-state power while preserving real-time feasibility, AMPC provides a better transient performance at a higher online computational cost and without guarantees of convergence to the steady-state optimum, and ESC offers a computationally inexpensive model-free baseline that may converge to locally optimal solutions. These results provide a practical reference for selecting wind farm control strategies and for designing scalable, real-time optimization methods.

Large language models (LLMs) are increasingly deployed as long-horizon agents with decision-making capacities. While LLMs can show ethical competence on dilemmas such as trolley problems, this competence may not translate to complex, agentic scenarios. We study this gap in Civilization V, a multiplayer game with a complex decision-making landscape including economy, diplomacy, technology, and military strategy. Starting from 130 high-tension LLM self-play episodes, in which an LLM player spontaneously escalated nuclear authorization, we replay them across 13 models with three prompt interventions: an ethical prompt naming nuclear harm, removal of the previous model's decision-making rationale, and high-stakes framing emphasizing real-world impacts. No interventions nor their combinations reliably eliminate emergent escalation. We identify three failure pathways: ethical reasoning that fails to surface without prompting, fails to appear even when prompted, or surfaces but fails to take effect when strategic counter-factors dominate. Evaluations of agentic models, therefore, must test whether ethical reasoning is spontaneously invoked and behaviorally effective in complex decision-making contexts, beyond whether it can be elicited in isolation.

Physical human-robot interaction (pHRI) demands simultaneous trajectory accuracy and compliant safety under unplanned contact. Classical impedance control incurs a nonzero steady-state position error under sustained human force -- the applied force divided by the task stiffness -- which integral action reduces only within a narrow stable-gain budget. We present a two-layer Impedance MPC that resolves this tension. Layer~1 analytically cancels gravity, Coriolis, and task-space inertia, reducing the residual plant to a configuration-independent double integrator with a constant state-transition matrix. Layer~2 solves a 30-variable convex QP at 100\,Hz, exploiting this constant structure so the free-response matrix is precomputed once; an augmented Kalman filter estimates the persistent disturbance state, giving a formal zero-steady-state-error guarantee. A null-space inverse-barrier potential and a task-space workspace projection enforce joint-limit safety across the tested workspace. On a 7-DOF Franka FR3, Impedance MPC with Kalman augmentation attains sub-0.05\,mm steady-state error versus 44.8\,mm for classical impedance (a $>$800-fold reduction) under a sustained 15\,N force, sub-millimeter tracking on four 3-D circles, and graceful robustness to measurement noise and inertial mismatch up to 30\%.

The emergence of large language model (LLM)-based agents and multi-agent systems has enabled a shift from narrow task automation to more autonomous decision-making. Despite progress in language generation, planning, tool use, and coordination, most agents still treat intelligence as prediction, optimization, and task completion. Human environments are social and normative, where people reason under bounded rationality, communicate in culturally situated language, and make decisions guided by values, beliefs, trust, and social norms. This survey argues that future AI agents, especially those acting on behalf of humans, must move beyond task competence toward human-centered capabilities. We review research across six areas: (1) evolution of intelligent agents, (2) human cognition and decision-making, (3) language, culture, and social context, (4) human values and belief systems, (5) human-agent collaboration, and (6) multi-agent coordination and modeling of human characteristics. We synthesize work from cognitive science, sociolinguistics, computational social science, and AI alignment, along with recent advances in LLM agents, cultural alignment benchmarks, preference learning, explainability, and agent societies. We identify a key gap: existing systems do not provide a unified framework integrating cognition, culture, values, and social behavior into autonomous agents. We conclude with directions for building culturally aware, value-aligned, cognitively grounded, and cooperative multi-agent systems.

Large Language Model (LLM) agent systems suffer from failures that occur without external triggers -- no injection, no adversarial input, no resource exhaustion. These silent failures -- unexpected deviations from intended behavior under normal conditions -- are routinely misattributed to bugs or configuration errors. Through systematic analysis of over 40,000 controlled trials and long-term production observations spanning 100,000+ agent interactions, we identify a common structural logic underlying these failures. Building on patterns observed in our experiments, we survey the global research literature on autonomous agent reliability and synthesize 22 intrinsic properties of LLM agent systems across six lifecycle layers: foundation semantics, inter-agent transmission, memory persistence, task execution, feedback correction, and systemic evolution. We demonstrate that whenever a sufficient subset of these properties co-exist, system entropy -- the measurable accumulation of disorder: loss of output consistency, task accuracy, and cross-session coherence -- increases monotonically with interaction rounds. We formalize this as the Entropy Principle: S(t) = S0 * e^(alpha * t), with alpha measured empirically across multiple architectures. We propose the PIG (Physical Integrity Gate) Engine with the ADE (Agent Delivery Engineering) protocol suite as an engineering countermeasure to entropy-driven disorder. Our findings establish silent failure not as a bug to be fixed but as a manifestation of Intelligence Entropy -- a physical constraint to be managed through deterministic governance. We argue that any engineering effort stabilizing the structure and order of agent systems participates in a unified mission: keeping intelligent systems reliable as they grow in scale and complexity.

Eukaryotic DNA replication must remain robust under thermal, chemical, and genotoxic stress despite large fluctuations in replication dynamics. Here, we develop a lattice-based stochastic Monte Carlo framework for whole-genome replication in Saccharomyces cerevisiae at single base-pair resolution, incorporating probabilistic origin firing, replication fork-speed distributions, and a time-dependent limiting factor that governs the availability of cellular replication resources. The model is benchmarked quantitatively against experimental replication profiles before being applied to stress conditions, and reproduces diverse replication stress responses using only two effective parameters. Importantly, the analysis reveals that replication fork-speed heterogeneity underlies the emergence of Erlang-distributed S-phase durations and rare, anomalously prolonged replication events observed experimentally in Escherichia coli and human cell lines, while predicting similar behavior in S. cerevisiae. The framework further predicts non-monotonic thermal behavior, power-law scaling under hydroxyurea stress, and total replication-time dynamics under diverse genotoxic conditions.

Affine formation control offers immense flexibility for coordinating multi-agent maneuvers, but guaranteeing the safety of agents under parametric uncertainties remains an open challenge. This paper proposes a novel safe affine formation control framework for second-order multi-agent systems by integrating Higher-Order Control Barrier Functions (HOCBFs) with Adaptive Dynamic Programming (ADP). We introduce a barrier-modulated control architecture that smoothly attenuates the nominal formation tracking objective when agents approach safety boundaries, preventing conflicting control inputs. Within this architecture, two distinct safety controllers are developed: (1) an analytical barrier-gradient repulsive controller that provides a computationally efficient, rigorous mathematical baseline, and (2) a data-driven optimal safety controller. The data-driven approach utilizes an actor-critic neural network to solve the Hamilton-Jacobi-Bellman (HJB) equation online, enabling optimal collision avoidance even in the presence of unknown system parameters. Using Nagumo's theorem and Lyapunov stability analysis, we formally prove that both controllers guarantee the forward invariance of the safe set ensuring absolute collision avoidance while maintaining Uniformly Ultimately Bounded (UUB) formation tracking errors. Finally, simulations validate the theoretical findings and demonstrate the robustness of the proposed controllers in dynamic obstacle avoidance scenarios.

Self-evolving agents improve by repeatedly proposing changes to their own prompts, skills, or workflows and keeping those that score higher on a small held-out set. Almost all effort has gone into the proposer that generates candidates; we argue the weak point is the acceptor, the rule that decides whether to commit a change. Applied hundreds of times against the same noisy dev estimate, the ubiquitous "keep it if the score went up" rule is uncontrolled adaptive multiple testing: the agent effectively p-hacks itself, accumulating false commits that make it churn and drift rather than improve. We recast committing as a sequential hypothesis test and propose PACE (Paired Anytime-valid Commit Evaluation), a training-free, anytime-valid commit gate. Each candidate is compared to the incumbent on identical instances and committed only when a testing-by-betting e-process accumulates decisive evidence, stopping early to save evaluations and controlling each candidate's false-commit probability at a user-set level even under optional stopping (a per-decision guarantee). On Qwen2.5 agents (0.5B-3B) self-evolving at the prompt level on GSM8K, SVAMP, and ARC-Challenge, greedy acceptance commits 30-42% false and 10-33% harmful edits when a genuine improvement is hidden among noisy proposals, while PACE commits the real one and essentially nothing else, matching greedy's held-out accuracy at sharply lower variance and about 18% lower evaluation cost. With no real gain available, greedy commits 13-21 spurious self-modifications per run (72-100% false) and degrades the most fragile agent by 4.9 points, while PACE holds at baseline. Reliability of self-evolution depends on the acceptor, not only on the proposer.

Multi-quadruped coordination has attracted increasing attention due to its enhanced payload capacity, broader contact coverage, and improved adaptability to challenging tasks. Existing methods for multi-quadruped manipulation typically focus on predefined or closed task families, often relying on multi-agent reinforcement learning (MARL) to train task-specific coordination policies. However, such methods struggle in open-ended continual learning settings, where tasks arrive sequentially and robots are expected to acquire new coordination skills while reusing previously learned ones without catastrophic forgetting. To address this challenge, we propose Conquer, a semantic skill-library framework that formulates continual multi-quadruped coordination as a retrieve-adapt-update process. First, to accommodate varying team sizes across tasks, we design a team-structured Self-Allies-Goal (SAG) backbone that supports variable-cardinality robot teams by explicitly modeling each robot's own state, teammate context, and task goal. For each incoming task, Conquer constructs a task-level semantic descriptor from pre-execution information and retrieves a relevant skill from the library for adaptation. After successful execution, Conquer updates the skill library by extracting trajectory-level semantic descriptors and organizing them according to semantic distance, thereby enabling continual skill accumulation and cross-task knowledge transfer. Simulation experiments show that Conquer achieves a final average success rate of 95.6%, demonstrating strong forward transfer and negligible catastrophic forgetting. Real-world rollouts on Unitree Go2 teams further validate the deployment feasibility of Conquer for practical multi-quadruped coordination. Simulation and real-robot demonstration videos are available at: https://conquer-project.pages.dev/.

Vision-Language-Action (VLA) policies are typically shipped as Python/PyTorch stacks that assume a workstation-class GPU, a mismatch for the hardware on which robots actually run. We present vla.cpp, a portable C++ inference runtime built on llama.cpp. To our knowledge, it is the first ggml-class engine to natively serve the flow-matching and diffusion VLA inference pattern, in which a cached vision-language prefix is consumed by a cross-attending action expert integrated over several solver steps. A single runtime serves seven architectures spanning five backbone and four action-head families behind one request/response protocol, with each model packaged as a self-contained bundle. On LIBERO-Object, the engine matches a state-of-the-art checkpoint to within one episode out of 200, and runs BitVLA at 100% success in 1.3 GiB of memory. The same bundle runs unchanged across three hardware tiers, from a consumer GPU down to an 8 GB embedded module. A cross-hardware roofline analysis shows that batch-1 VLA inference is compute-bound, so utilization rather than bandwidth is the deployment lever; an IMMA ladder GEMM derived from this analysis cuts BitVLA per-step latency by 4.5x. We then frame an on-robot stress test on an ALOHA arm that isolates the latency constraint under which a learned VLA must replan against a moving target on the hardware it was trained for. Code, demo videos, and the reproducible benchmark scaffold are available at https://fai-modelopt-tech.github.io/vla-cpp.github.io/.

AI agents increasingly turn past experience into reusable artifacts such as code, workflows, and procedural memories. Reuse can improve efficiency, but it also creates a lifecycle reliability problem: artifacts that succeed once may fail under environment drift, underspecified tasks, or changing task distributions, especially in web automation. We introduce SKILL.nb, a framework for governing reusable agent workflows with evidence-calibrated lifecycle policies. SKILL.nb uses selective formalization: execution evidence decides which workflow steps should become executable code, which should remain natural-language guided, and when those choices should be revised. Workflows are stored as auditable, versioned notebooks that interleave natural-language guidance, multi-language executable cells, validation gates, fallback paths, and multimodal evidence such as outputs, screenshots, and error traces. At runtime, gate-conditioned execution lets each step run code when its gates validate, or fall back locally when drift invalidates the executable realization. On WebArena-Verified, SKILL.nb achieves 53.7% single-round success, improving over the strongest baseline by 3.9 percentage points. Across three re-executions, it retains 91.7% of initially successful tasks, 15.5 points above the next best method. Under bounded repair, it recovers 72.9% of subsequent failures while limiting post-repair regressions to 4.2%, compared with 15.0% to 17.0% for persistent baselines. It also leads on Mind2Web cross-website and cross-domain splits. In a GitLab migration test, SKILL.nb preserves performance when reusing frozen state learned on GitLab 15.7, with frozen-versus-fresh target-version gaps of -1.7 points on GitLab 16.11 and +0.6 points on GitLab 18.9. These results identify lifecycle governance and gate-conditioned execution as reliability axes beyond one-shot task success.

Agentic tutoring systems introduce a coordination challenge: multiple agents may propose different but reasonable interventions, yet only one response can be delivered to the learner. In this paper, we study how voting protocols shape cooperation among four role-constrained pedagogical agents responsible for scaffolding, misconception, motivation, and metacognition. We compare four voting protocols -- simple, ranked, cumulative, and approval voting -- across two simulated tutoring environments on SciQ and HumanEval benchmarks. Rather than using voting as a simple aggregation step, we use it to analyze how collective decision rules shape coordination under partial pedagogical conflict. Across 1,200 simulated interactions, we find that agent deliberation and voting protocol type frequently change which response ultimately wins, showing that both meaningfully shape the collective decision. Different voting rules also produce distinct coordination behaviors, and even brief tutoring turns show measurable learning gains in simulated students. Overall, we show that protocol choice is associated with distinct coordination patterns among role-specialized pedagogical agents.

As large language model (LLM) agents are integrated into autonomous cloud operations, distributed systems face a semantic reliability problem: proposer agents can generate production mutations, such as modifying IAM policies, opening firewall security groups, or executing data exports, that are syntactically valid and statically authorized but operationally unsafe. Classical distributed consensus protocols replicate deterministic state transitions but do not evaluate the safety of the proposed intent. To address this gap, we introduce Semantic Quorum Assurance (SQA), a control-plane primitive for governing non-deterministic agentic infrastructure. SQA represents proposals as declarative execution contracts bound to cryptographic evidence chains and routes them to a diverse panel of read-only, sandboxed validator agents. SQA aggregates their judgments under a risk-adaptive quorum predicate that enforces model and archetype diversity, adjusts weights based on calibrated assurance scores, and respects archetype-specific vetoes. Admitted proposals execute only through a sovereign execution gate. We instantiate SQA in a cloud-native control plane and formalize a correlated cognitive failure model for non-deterministic validators. On 500 infrastructure-inspired mutation scenarios, with safety results reported on held-out safe/unsafe trials excluding ambiguous scenarios, SQA reduces unsafe approval from 18.5% for single-agent validation to 0.3% while adding median validation latency of 1.45--4.12 seconds across the studied risk buckets.

Understanding how educational social dynamics evolve is critical for informing effective educational policies and counterfactual interventions. However, traditional methods face a fundamental dilemma: observational studies often lack causal power, while controlled experiments are frequently constrained by ethical concerns. Although LLM-based multi-agent simulations offer a scalable in silico alternative, existing approaches remain limited by weak psychological grounding and insufficient measurement of latent psychological states. To address this, we introduce EduMirror, a multi-agent simulator for the scientific study of educational social dynamics. We provide configurable education-oriented agent forms, including value-driven agents grounded in psychological needs and social value orientation, together with a dual-track measurement protocol for quantifying observable behaviors and latent psychological states. We validate the realism and usability of EduMirror through case studies on school bullying and group cooperation, as well as broader evaluations across diverse educational scenarios. The results show that EduMirror generates educational social dynamics that are realistic, theory-consistent, and measurable by empirical criteria. These properties enable structured in silico educational research, providing a computational tool for hypothesis testing and counterfactual intervention analysis in educational science. Project page: https://edumirror.net.

Regulatory review of advanced nuclear reactor designs routinely spans more than three years and consumes hundreds of millions of dollars in combined regulator and applicant labor. We present the Regulatory Context Protocol (RCP), an Agent-to-Agent communication standard that replaces the formal human-to-human pipeline between regulators and applicants with a structured, auditable agentic channel, while preserving human oversight at safety-significant decision points. The protocol is calibrated against an analysis of 1,236 documents from U.S. Nuclear Regulatory Commission advanced reactor dockets and demonstrated with a working multi-agent pilot. Against an 89M USD, 42-month Reconstructed Baseline, RCP cuts costs by 50-77 percent (21M-44M USD) and timelines by 65 percent (15 months). Without a shared protocol, Standalone Agents reach only 54M-74M USD and 21 months. The residual cost-and-time gap is structural, not algorithmic: it traces to the inter-organizational pipeline that only an agent-to-agent standard can compress. The same bottleneck - formal multi-party review under strict auditability requirements - characterizes pharmaceutical approvals, environmental permitting, financial supervision, and aviation certification. The US regulatory paperwork burden carries a 426.5 billion USD annual opportunity cost; replicated broadly, the projected 50-77 percent reduction implies savings on the order of 210-330 billion USD per year - approaching 1 percent of US GDP.

We measure how well current large language models coordinate as multiple agents sharing a common resource, using the dining philosophers problem as a clean test bed. Across 630 episodes spanning seven models and three philosopher counts, four frontier closed-source systems reach mean reward 0.45 to 0.87 and Mistral-Small 24B reaches 0.83 to 0.99, while Qwen3-14B reaches 0.13 to 0.35. We then ask whether group relative policy optimization (GRPO) on rollouts from the task itself can close the gap and find that it cannot: a Welch's t-test on per-episode reward at five philosophers gives p = 0.66 and a Hedges' g of -0.11, with no statistically significant change at ten or fifteen philosophers either. Two further observations qualify the result. The training reward of both 8B and 14B runs peaked at step nine and then declined, so the default saved checkpoint at step 15 is strictly worse than several earlier ones. The four-term reward we use admits a degenerate maximum at zero actions, which DeepSeek-R1-Distill-Qwen-7B and Mistral-Small 24B at five philosophers both inhabit, with mean reward 1.0 and 0.83 respectively at zero meals. The bottleneck for an open-weight 14B model on multi-agent coordination is not training compute but training methodology: reward shaping that does not collapse to a no-action maximum, checkpoint discipline that does not depend on the final step, and curriculum across problem scales.