Physical AI Brief

We introduce Hydra-0, a generalist world model conditioned on action flow, which represents robot actions as pixel motion. This shared visual interface enables generalist world modeling and control by learning action consequences across embodiments, tasks, environments, and video-generation backbones. Our best configuration achieves 90.4% lower robot-motion error and 60.2% lower object-motion error than our action-conditioned baseline, while supporting zero-shot composition and data-efficient adaptation. On the RoboLab benchmark, Hydra-0 achieves a Pearson correlation of r=0.96 between replayed and reference success rates. Finally, we uncover an emergent inverse mode of this interface: a world action model that predicts compatible robot motion from desired object flow transferred from a human demonstration. A trained action head maps the resulting latent features to executable actions without requiring task-specific expert robot demonstrations. Together, these results demonstrate the potential of action flow as a shared control interface connecting heterogeneous training data, open-loop policy evaluation, and robot control.

Purpose: To develop and evaluate a locally deployed multi-agent AI system for radiology report structuring and quality assurance. Materials and Methods: This retrospective study included 638 radiology reports from CT examinations of the chest, abdomen, and pelvis dictated by 15 board-certified radiologists in 2023 and 2024. A multi-agent AI pipeline was developed to perform report structuring and quality assurance (QA). The system structured the report into standardized anatomical sections at the sentence level using regex rules and local large language models. It also detected mismatches between the Findings and Impression sections, or within sections; gender-anatomy conflicts; and undocumented communication of critical findings. Two board-certified radiologists independently evaluated a 45-report subset. Results: The multi-agent system structured the Findings sections of all reports (22,270 sentences) into a predefined anatomical format while retaining the original report content. The system flagged 90 (14.1%) reports, most commonly for section mismatches (80 reports, 12.5%). In the radiologist evaluation, both reviewers agreed that 31 (69%) were correctly restructured, 2 reports (4%) were incorrectly restructured, and disagreed on the remaining 12 reports (27%). Both reviewers agreed that no clinically important information was omitted and no fabricated content was introduced. Overall QA performance was rated as "excellent" or "good" in 84% of the evaluated reports, with the remaining reports rated as "fair". Conclusion: A locally deployed multi-agent AI system combined radiology report structuring and quality assurance within a single workflow. The system demonstrated favorable performance in radiologist evaluation. Such systems may support standardization of reporting and quality assurance in radiology practice.

Memory-based self-improving agents--those that learn from an online stream of tasks and improve over time by maintaining a textual memory bank--have shown great promise in recent literature. However, the reliability aspects of these methods have been critically overlooked. In this work, we conduct a comprehensive re-evaluation of two memory-based methods, broadening the scope of evaluation along two axes: (1) including multiple runs to quantify variance, and (2) randomly shuffling the tasks to investigate the effect of task order. Through these experiments, we make two observations that expose the fragility of current methods: First, agent evaluation is inherently noisy in complex environments and on multi-step tasks, and stacking a self-improving loop on top can further amplify this noise. Second, the agent's improvement is highly dependent on task order. Prior works often adopt default orderings that impose an implicit curriculum, acting as a hidden prerequisite for success. To better understand this fragility, we manually examine the agents' memory and hypothesize that task and environment underspecification contribute to this fragility. We validate this hypothesis by incorporating information that enables better specification, such as detailed rubrics and environment feedback, into the memory construction process. While this added information partially closes the performance degradation in previous experiments, significant gaps still remain, suggesting that other uncharacterized factors contribute to this fragility. Looking ahead, our work advocates for more rigorous evaluation protocols for self-improving agents by reporting results across multiple runs and stress-testing them under challenging conditions. Moreover, our findings on underspecification call for systems and interfaces that enable effective human oversight, preventing agents from failing in unforeseeable ways.

High-resolution image editing is increasingly demanded in professional workflows, yet existing diffusion-based models remain constrained to resolutions below 1K due to quadratic attention complexity and prohibitive memory requirements. A prevalent workaround employs a two-stage pipeline: editing at low resolution followed by independent super-resolution. However, this approach suffers from two critical issues: information divergence, where hallucinated details contradict the original high-resolution (HR) source, and texture degradation, manifesting as over-smoothed or over-sharpened artifacts. We propose EditBridge, a diffusion bridge framework for efficient ultra high-resolution editing. Unlike conventional diffusion that regenerates from noise, we formulate refinement as structured data-to-data translation from the low-resolution (LR) edited result to its HR counterpart, explicitly conditioned on the original HR source to preserve authentic details. To efficiently incorporate HR source guidance, we introduce a prior-guided block-wise sparse attention mechanism that exploits semantic correspondence from first-stage editing to constrain cross-image interactions to spatially aligned regions, significantly reducing computational overhead. Extensive experiments demonstrate that EditBridge achieves high-fidelity editing with superior perceptual quality at resolutions up to 4K, delivering 3.6--8.4$\times$ speedup at 2K and enabling practical 4K editing in 61 seconds.

Language model tokenizers are typically selected with minimal evaluation, despite the fact that their design choices directly impact model capabilities. This can be partly attributed to a limited understanding of which tokenizer properties affect which aspects of downstream performance. We introduce TokEval, a framework of tokenizer evaluation metrics that goes beyond standard measures like fertility and compression rate to capture linguistically and structurally meaningful properties, e.g., UTF-8 character boundary integrity and digit place-value boundary alignment for mathematics. To validate whether these metrics are predictive of downstream model performance, we conduct controlled language model pretraining experiments, varying solely the tokenizers' training data mixture, pretokenization strategy, and training algorithm. We evaluate the resulting models on bits-per-byte (a tokenizer-agnostic version of perplexity) and several benchmarks, spanning linguistic understanding, mathematical reasoning, and code generation. Our experiments suggest that different intrinsic properties have different impacts on model abilities: information-theoretic metrics predict language modeling abilities (Spearman rho up to 0.80), while structure-sensitive metrics, such as those measuring digit and line-break handling, correlate with task accuracy. We hope TokEval enables more principled tokenizer evaluation, replacing pretraining sweeps with intrinsic measurement wherever the two agree.

We give a two-player zero-sum repeated game between a learner and nature whose value identity generates Bayesian updating and an exact accounting of exponential-weights regret at once, and supplies the comparator-class variational form that a wide class of concentration phenomena share. The terminal payoff is the most a comparator can gain at fixed relative entropy from the prior, and the one-step constraint is an information budget on nature's move under the learner's mixed action. With the learner's move otherwise unrestricted, Gibbs/Bayes weights emerge as its unique Bellman equalizer -- the mixed action that makes the per-round loss independent of which direction nature moves -- with log-partition functions playing the role of value functions. The regret decomposes exactly into three parts: a per-round information loss reflecting the variation in observed outcomes, an additive retempering drift that accounts exactly for any change of measurement scale between rounds, and the information the comparator carries relative to the prior. The variance and bounded-range proxies that drive standard regret bounds are looser relaxations of this decomposition, which holds generally and governs them all. Both players' strategies are read off from the decomposition term by term, and repeated play yields an information-theoretic ledger of self-play in place of the usual quadratic-variation surrogate. The same comparator-class geometry accounts for the classical large-deviation bounds, and methods across bandits, posterior sampling, aggregation, and boosting are specializations of the one regret decomposition.

Resistive random-access memory (ReRAM) is a promising future nonvolatile memory technology. Most ReRAM exhibit a fundamental tradeoff: filament-type ReRAM provides long data retention but suffers from poor uniformity and high switching current, whereas nonfilamentary ReRAM shows lower-current, area-dependent switching but generally poor retention. No two-terminal device has been able to overcome this tradeoff. In this work, we present a two-terminal Ta2O5/ TaOX resistive memory cell that achieves both nonfilamentary switching and long retention. Electrical measurements and composition depth profile show that the switching is not confined to a single filament but is instead uniform across the entire switching region. Despite the nonfilamentary switching, this device can retain information for over 22 hours at 190 °C, which is comparable to the best filamentary devices. We propose that this long retention arises from composition phase separation in amorphous tantalum oxide. Our work shows that the fundamental tradeoff between information retention and switching uniformity can be overcome, and thereby provides a pathway toward more uniform and reliable oxide memory devices.

Autonomous LLM agents that converse on a user's behalf are an emerging design pattern in matching platforms, yet their viability depends on a condition rarely examined: users must accept not only delegating conversation to an agent, but also receiving agent-mediated communication from others. We study this condition using two large-scale surveys of active users of a major dating platform (N=2,894 on generative profile features; N=2,617 on autonomous conversational agents, fielded in two languages). We develop a latent-variable measurement model of agent receptivity based on graded response models with latent regression, and show via model comparison that willingness to send and willingness to receive agent communication are distinct constructs: highly correlated (rho=0.92) but separable (Delta BIC=52), with partial measurement invariance across languages. The model quantifies a systematic delegation asymmetry: deploying one's own agent requires far lower receptivity (threshold -0.38) than engaging a counterpart's agent (+0.32; full engagement +1.39), and mean deployment propensity exceeds engagement propensity roughly threefold. Under a random-pairing counterfactual derived from stated receptivity, only 4-13% of directed dyads combine agent deployment with receiver engagement, with a pronounced gender-directional imbalance. Design counterfactuals quantify the levers: a reciprocity requirement cuts interaction volume by half or more by excluding nearly two-thirds of would-be deployment, while routing agent contacts on receive receptivity triples per-contact engagement, a lift that survives out-of-sample validation with the target item held out (AUC 0.88, 3.1x quartile lift under respondent-level cross-validation). We discuss implications for agentic recommender design, including disclosure, opt-in mechanics, and receptivity-aware matchmaking.

AI agents increasingly perform knowledge work (i.e., produce and modify persistent digital artifacts such as code repositories, documents, spreadsheets, slides, reports), yet the parsed views they search, the native files they edit, the changes they review, and the artifacts they submit can refer to different versions of the same work product. We formulate this as a workspace-state contract: every view should be explicitly tied to a version of the evolving workspace state. Coding agents partly address this need through repository contracts for search, diffs, and tests, whereas an analogous contract is less explicit for PDFs, spreadsheets, slides, notebooks, and mixed-format project folders. We propose StagedWorkspace, a versioned workspace for knowledge-work agents. The workspace binds parsed records and review diffs to content hashes of the native files as they change. In fixed-harness ablations on OfficeQA Pro and APEX-Agents, dual parsed/native access has the highest point estimate for every tested model; relative to the more limiting single view, it improves OfficeQA Pass@1 by 8.3-12.1 points and APEX mean rubric score by 4.7-9.2 points. SW-AGENT scores 63.9% with Gemini 3.1 Pro on OfficeQA and 42.1 with GPT-5.4 Nano on APEX, compared with published same-model scores of 29.3% and 25.5, respectively. A paired review-axis ablation on 57 file-editing tasks further finds higher observed scores when diffs are visible. These results identify workspace state as an experimental variable in knowledge-work agents and motivate benchmarks that score evidence, staged edits, and submitted artifacts as explicit state transitions.

Traffic elements such as traffic lights and road signs play a fundamental role in human driving decisions and should naturally influence end-to-end driving performance. However, existing end-to-end driving research predominantly focuses on dynamic road participants (e.g., vehicles and pedestrians), while the role of traffic elements remains largely unexplored. The community still lacks a systematic study quantifying their impact, largely because public datasets rarely provide structured traffic-element annotations and modern driving systems vary widely in architecture and training paradigm. In this work, we present the first systematic investigation of traffic element awareness for end-to-end autonomous driving. We construct a unified research infrastructure by augmenting multiple public driving datasets with comprehensive traffic-element annotations. To support diverse model families, we adopt a minimal and universal integration design that incorporates traffic-element signals into existing pipelines in a plug-and-play manner with negligible architectural modification. We evaluate this design across modern paradigms, including perception-prediction-planning pipelines, vision-language-action models (VLA), regression-based planners, diffusion-based policies, and trajectory-scoring frameworks, on nuScenes, NAVSIM-v1, NAVSIM-v2, and Bench2Drive. Across all paradigms and datasets, this simple integration consistently improves driving performance, demonstrating that traffic element awareness provides a robust and generalizable signal for end-to-end driving systems. Notably, on the challenging NAVSIM-v2 benchmark, our approach significantly improves state-of-the-art architectures and data pipelines, establishing a new state of the art.

Academic surveys play a central role in organizing rapidly expanding scholarly literature, yet their construction requires extensive paper analysis, coherent knowledge organization, fine-grained citation support, and reliable manuscript assembly. Existing Deep Research and automated survey generation systems address parts of this process, but typically do not coordinate paper understanding, literature organization, evidence-grounded drafting, and manuscript validation through a shared, revisable state. We introduce DAS, a stateful agentic framework for generating publication-oriented academic surveys. Its key idea is to separate reusable paper analysis from topic-specific manuscript construction. DAS builds on DAS-2M, a dynamically updated metadata lake containing survey-oriented representations of approximately two million papers. Its agents maintain explicit literature, organization, writing, and finalization states through candidate-grounded taxonomy planning, reverse paper-to-section routing, and hierarchical claim and citation planning. Semantic review reactivates only the affected writing states for repair and reevaluation, forming a scoped closed loop with deterministic validation. We further introduce DAS-Bench, a 30-topic benchmark, together with DAS-Eval, which assesses scholarly citation quality, taxonomic synthesis, hierarchical discourse, and manuscript assembly reliability through 16 criteria. Among systems evaluated on all 30 topics, DAS achieves the highest average in all four dimensions, with an overall score of 4.34 compared with 4.03 for the strongest competitor, and the same ordering is preserved on the matched 21-topic CS subset. Blinded expert evaluation further prefers DAS to Naive RAG on 27 of 30 topics and to AutoSurvey on 19 of 21 shared CS topics. The project page is available at https://zhikaixu24.github.io/projects/DAS/.

Initialization-free bundle adjustment (InitFree BA) aims to recover camera poses and scene structure directly from image observations, avoiding the geometric initialization stages of conventional structure-from-motion pipelines. Recent methods based on Object-Space Error (OSE) formulations and Variable Projection (VarPro) show encouraging optimization behavior from random camera configurations. However, existing evaluations primarily measure optimization success, leaving unclear whether a low OSE objective yields a valid metric 3D reconstruction. We revisit InitFree BA experimentally through a unified evaluation framework combining a C++ implementation of existing OSE formulations with a Blender-based dataset generator providing exact ground truth and controlled camera configurations and observation densities. Our experiments reveal a previously overlooked optimization--reconstruction gap: projective solutions with similarly low OSE values can lead to substantially different Euclidean reconstructions after metric upgrade. We identify initialization priors, landmark observation density, and metric-upgrade stability as key factors governing reconstruction success. Overall, our results suggest that the main challenge of InitFree BA is not merely minimizing OSE objectives, but obtaining projective reconstructions that admit reliable metric upgrade. We believe that the proposed benchmark, implementation, and analysis establish stronger experimental foundations for future research on initialization-free bundle adjustment, a problem largely unexplored within the computer vision community. Project page is available at https://github.com/simonwebertum/InitFreeBA.git.

Large-scale, high-dimensional tabular regression remains challenging: tree-based models are robust but lack end-to-end representation learning, while deep models enable flexible feature learning but often incur costly interaction modeling and sensitivity to noisy or redundant features. We propose TabNSM, a scalable regression framework that extends our earlier sparse-attention and mixer architectures. At its core, the Adaptive Sparse Interaction Module (ASIM) integrates foreground feature discovery, sparse local interaction encoding, and Feature-Token Mixing, providing near-linear complexity under fixed sparse configurations. For regression, TabNSM introduces three complementary components: a Multi-Stage Regression Head for progressive prediction refinement; GridLoss, an ordinal-aware soft-binning objective that incorporates target structure into representation learning; and RISE (Reweighted Instance Sampling by Error), a difficulty-aware sampling strategy based on loss-quantile bins. Across nine real-world regression benchmarks, TabNSM delivers strong predictive performance and practical scalability, with particularly consistent gains on high-dimensional and heterogeneous datasets. These results demonstrate that selective interaction modeling, structured regression supervision, and difficulty-aware sampling provide an effective and scalable approach to deep tabular regression.

WEASEL 2.0 is a dictionary-based time series classifier that combines dilated sliding windows with a randomised hyperparameter ensemble and a fixed-size dense feature representation. Two of its hyperparameter choices, the maximum ensemble size and the maximum window size, are specified by simple thresholding rules whose chosen thresholds are not empirically justified in the original paper. In this work we reproduce WEASEL 2.0 on 114 UCR datasets, achieving a mean accuracy of 0.865 and median of 0.928, closely matching the published values (Wilcoxon signed-rank, p = 0.655). We then test the sensitivity of four design choices: the downstream classifier, the absence of feature weighting, the maximum window-size rule, and the maximum ensemble-size rule. The first three are robust to perturbation. The fourth is over-provisioned for long-series datasets, motivating an adaptive rule that sets the maximum ensemble size from series length and number of classes. Evaluated on fixed-length datasets, the adaptive rule reduces peak fit memory by a median of 37 MB (mean 395 MB) and fit time by a median of 0.4 s (mean 4 s), with a median accuracy change of 0% (mean -0.11%). Memory and time savings concentrate on long-series datasets where the original rule allocates the largest ensemble size.

Reasoning in LLMs is overwhelmingly studied in domains that provide a model with rules: mathematics and code. Linguistic puzzles invert this: the solver must first discover the system before reasoning within it. We present the IOL-AI Challenge, an open-science competition run on the unseen problems of the International Linguistics Olympiad (IOL) 2026 Individual Contest, evaluated both automatically and, for the first time, by members of the official IOL Jury under the same rubrics applied to human contestants. The challenge drew 731 submissions from 46 teams under a strict compute budget (one T4, 30 mins). We additionally benchmark 15 unconstrained frontier and open models, with Claude Opus 4.8 earning a jury score equivalent to a gold medal, while both resource-constrained systems we submitted for jury grading scored in the range of the bottom 5% of contestants. Capability was not determined by scale: 14B submissions outperform models twice their size, and gains come from decoding and output-handling rather than model capacity. We also found that automatic metrics rank systems exactly as the jury does, but compress the scale, upscoring weak systems by ~13 points and understating strong ones. Our analysis shows that while frontier models might have prior knowledge about some of the problem languages, it does not significantly help them solve the linguistic reasoning tasks, leaving linguistic reasoning as a strong benchmarking proxy for generalizable reasoning skills.

Answering questions accurately and efficiently in embodied scenarios presents significant challenges due to limited computational and memory resources for Vision Language Model (VLM) inference. Existing methods adopt visual search key frame retrieval method to select critical question-related key frames for VLM input. However, visual search methods are inefficient because they require visual search among thousands of video frames for each individual user query. In this work, we propose a memory tree guided key frame selection paradigm for efficient 3D question answering in embodied scenarios. Our method leverages a compact and reusable 3D scene representation, termed MemTree3D, which supports real-time online construction leveraging camera 6-DoF poses. MemTree3D captures multi-level 3D scene information, enabling a Large Language Model to efficiently query and retrieve question-relevant key frames through our scoring-based frame selection without reprocessing the entire video stream. On OpenEQA, our method improves the LLM-Match of GPT-4o by 17.4%, LLaVA-OneVision-7B by 5.8%, outperforms existing visual search methods. Our code is available at https://github.com/hsiangwei0903/MemTree3D

Using LLMs as judges has become standard practice for evaluating model outputs at scale. This is particularly common for subjective, open-ended tasks such as assessing helpfulness or alignment, where no single reference answer exists. However, objective tasks introduce a distinct reliability challenge for reference-free LLM judging. In the absence of a reference answer, the judge evaluates factual correctness either through its parametric knowledge or through tool augmentation. Although the former enables efficient evaluation, the judge may hallucinate or lack sufficient evidence for its verdict. Conversely, tool augmentation can provide additional evidence but introduces extra computational cost and requires an appropriate mechanism to determine when and how that evidence should be used reliably. More importantly, neither approach alone provides formal control over the risk of accepted verdicts or guarantees their reliability at a specified level. We propose a risk-controlled framework that calibrates uncertainty thresholds on a held-out set so that the false discovery rate among accepted verdicts remains below a user-specified level~$α$ with high probability, using finite-sample Clopper--Pearson intervals. When the parametric mode is not sufficiently confident, the instance is routed to a retrieval-augmented mode, where the judge gathers web evidence and re-evaluates the instance under a second calibrated threshold. The finite-sample guarantee carries over to this two-threshold routing without additional assumptions. Across open-domain QA benchmarks and judges of varying scales, the framework maintains the target error rate while achieving substantially higher coverage than single-mode baselines.

Authorship verification (AV) assumes that an author's writing style remains sufficiently stable to distinguish it from that of other writers. In practice, however, this assumption is challenged by distribution shifts caused by changes in genre, time, and AI-assisted writing. Existing AV benchmarks typically study these factors in isolation and focus predominantly on English, limiting our understanding of model robustness under realistic conditions. We introduce AVShift, the first German benchmark for systematically evaluating AV under multiple distribution shifts. AVShift comprises over 150,000 text pairs spanning three genres and 21 years, enabling controlled evaluation of cross-genre, temporal, and AI-era shifts within a unified framework. We benchmark representative feature-based, embedding-based, and LLM-based approaches. Our experiments show that fine-tuned LLMs generalize best across genres and benefit substantially from stylistically diverse training data. We further demonstrate that temporal drift is one of the strongest factors affecting AV, with performance degrading significantly as the time gap between documents increases. In contrast, we find no evidence of a measurable AI-era distribution shift within AVShift. Finally, our feature analysis reveals stylistic features that remain stable across genres, while their relative importance varies depending on the specific genre transition. We release AVShift and our code for future research.

Programmatic representations provide a compelling paradigm for 3D content creation, enabling fine-grained edits, interpretability, and explicit structural control. Yet, agentic workflows that rely on large language models (LLMs) to author 3D programs remain brittle, often failing to translate high-level intent into consistent low-level geometry. We attribute this fragility to a mismatch between existing programmatic interfaces and the reasoning strengths of LLMs, which favor semantic structure and spatial relations over fragile numeric choices. In this paper, we jointly design an Agent-centric Domain-Specific Language (aDSL) and a role-specialized multi-agent system to close this gap. aDSL bridges semantic logic and geometric constraints by emphasizing composability and spatial reasoning; it enables agents to manipulate geometry through relational operators instead of brittle absolute coordinates. Building on aDSL, our training-free multi-agent system follows a Plan-Execute-Critic loop to decompose requests, synthesize code, and iteratively repair errors and constraint violations using execution feedback. Experiments show that this co-design improves robustness, controllability, and faithfulness to user intent. Our method outperforms prior LLM-based baselines on text-to-shape and image-to-shape tasks while preserving explicit structure, editability, and interpretability. It also enables downstream applications such as articulated object creation and structured scene composition. Our code is available at https://github.com/sig-pku/aDSL.

Sparse attention mechanisms, which score all token pairs but propagate only the strongest, now underpin the most efficient Transformers for lightweight image super-resolution. This paper observes that sparsification changes what it means to improve such a network. A dense attention layer has one place where representation quality matters: the aggregation of attended features. A sparse layer has two, because the top-k operator first decides which tokens survive and only then decides what to do with them, and a token discarded at the selection stage cannot be recovered downstream. Selection quality and aggregation quality are therefore separable targets, addressed by modules placed before and after the attention respectively. We test this by pairing a dual-branch spatial enhancement on the input of a progressive focused attention with a wavelet-domain modulation on its output, forming SFMformer. Measuring each module alone and jointly over all fifteen benchmark-scale pairs, we find their gains are not additive: the joint gain exceeds the sum of the individual gains on nine pairs, and the sign of the discrepancy is predicted by how much the weaker module contributes on its own (r = -0.72), so the two compound when they relieve different constraints and overlap when they relieve the same one. Enabling spectral modulation once per block rather than once per layer retains the effect at roughly one-sixth of its cost, keeping the model below one million parameters at every scale. SFMformer ranks first on 28 of 30 PSNR/SSIM entries across five benchmarks and three upscaling factors. We report the cases where the pairing does not help, and deploy the model on a Raspberry Pi 5 to confirm the design is practical under tight resource budgets.

Online log anomaly detection is critical for maintaining the reliability of large-scale computing systems. Although recent language model-based log anomaly detectors achieve strong detection performance, their confidence estimates remain poorly calibrated. We show that these detectors frequently assign excessive confidence to incorrect predictions, particularly for anomalous logs under severe class imbalance. Moreover, confidence on erroneous predictions remains persistently high even when conventional calibration metrics indicate good calibration, creating a critical reliability gap for operational monitoring systems. To address this issue, we propose Log Reconstruction and Distance (LoRD), a lightweight post-hoc calibration framework for reliable log anomaly detection. LoRD learns prediction-route-specific reliability models from latent representations of correctly classified validation samples and estimates prediction reliability through route-wise reconstruction distances. Based on the estimated reliability, LoRD selectively recalibrates high-risk predictions to suppress overconfident errors while preserving reliable predictions. Extensive experiments on four large-scale log benchmark datasets and multiple language model-based detectors demonstrate that LoRD consistently improves confidence reliability and substantially reduces overconfident anomaly-related errors without sacrificing anomaly detection performance.

Recent progress in robotic learning has been fueled by large-scale datasets collected in everyday environments. However, most existing datasets emphasize short-horizon, low-contact tasks such as pick-and-place, and therefore do not capture the precision control, force/torque or tactile regulation, and multimodal feedback required for industrial assembly. To address this gap, we introduce PRISM, a large-scale multimodal dataset for contact-rich industrial operations. The dataset spans more than 25 manipulation tasks (e.g., electronic components plug/unplug, conveyor-based sorting) and covers diverse mechanical constraints. PRISM includes more than 5,000 trajectories totaling 45 hours of teleoperated demonstrations, recorded using synchronized multi-view RGB-D, force/torque, tactile, and robot-state measurements. In contrast to datasets collected in household or laboratory settings, PRISM provides a realistic benchmark for multimodal perception and control under high-precision industrial constraints, and serves as a foundation for contact-rich, generalizable manipulation in real-world manufacturing environments. The dataset is open-sourced at: https://tengbo-yu.github.io/PRISM/

State-of-the-art model-based reinforcement learning methods learn neural world models that allow policy improvement by planning in a latent space, without assumptions on the structure of the underlying environment. While expressive, these models are generally task-dependent: they learn uninterpretable latent representations that are tied to the training task and thus hard to generalize to new tasks. In this work, we present a novel world model formulation where the reward prediction only depends on a subset of structured, symbolic components of the whole latent state. Decoupling observation reconstruction and reward prediction allows us to learn world models that can adapt zero-shot, i.e. without further environment interactions, to new reward functions defined over the same symbolic state space. We discuss the main advantages and challenges of learning these neurosymbolic world models and demonstrate the strong generalisation properties of our approach over purely neural methods.

In the Code World Model paradigm an LLM synthesizes an executable world model that a classical planner searches, and the model is accepted when it reproduces sampled transitions. We ask what that acceptance certifies in continuous control. We define the pipeline's danger as an expected risk and isolate its exact factor: the probability that N i.i.d. gate rollouts all miss a critical event of probability r is exactly (1-r)^N; an independent acceptance sample adds its budget to the exponent. On three hybrid instruments the accepted mode-blind model is exploited: the planner is pinned at the mode boundary at a regret of nearly the whole attainable return. We prove a localization budget, valid at boundary points: models with Lipschitz constant at most L differing by eta at a point disagree above tolerance eps on a region of volume at least kappa((eta-eps)/L)^(d+m); the discontinuous reset modes studied pay no such budget. With real LLM synthesis, GPT-5.x repairs an omitted 1D clamp in 105 of 111 mode-containing draws -- every attempt exact on 50 of 56 instrument-stream blocks (95% CI [0.781, 0.960]). On 2D regions no artifact recovers the rule (0/156); eight targeted interventions leave the failure in place, and positive controls locate it: a located rule is not induced, while given form and location the constants follow exactly. A version-space certificate proves identification is class-relative: at the widest dose the declared fit succeeds in 20/20 blocks and every sample-consistent circle is within tolerance in 18/20. We prove a class of entry rules exactly consistent with every sample yet harmless at play, so identifiability is a measurable property of the instrument. Re-scoring all 1034 artifacts on independent samples confirms acceptance certifies sample consistency and no more: where the gate is provably informative it covers about two percent of the exploited planner's queries.

Large Language Models (LLMs) demonstrate remarkable multi-hop reasoning capabilities over long contexts, yet the internal mechanisms enabling these distant cognitive leaps remain poorly understood. Traditional attention-based interpretability often fails to capture true semantic proximity due to routing artifacts like attention sinks. In this paper, we bypass attention weights to directly analyze the dynamic geometry of the hidden state manifold, proving that deep LLM latent spaces natively organize into Small-World networks. By sparsifying the continuous similarity matrices of long-context representations into unweighted graphs, we trace the connectivity between highly disjoint semantic anchors across two distinct architectures. Our findings reveal a sharp topological phase transition: while early syntactic layers remain entirely fractured, deep reasoning layers abruptly compress massive conceptual distances into highly navigable pathways strictly bounded by the "Six Degrees of Separation" limit (=< 6 semantic hops). Furthermore, we demonstrate the practical efficacy of this framework by applying it to zero-shot hallucination detection within Retrieval-Augmented Generation (RAG) using the RAGognize dataset. We show that factually grounded generations maintain structural integrity with their source context (approximately 3 hops), whereas hallucinations induce severe topological collapse. Ultimately, this work mathematically formalizes how transformers execute abstract reasoning and provides a novel, strictly geometric signature for evaluating factual reliability.

Machine unlearning removes the influence of specific data from a trained model. However, most methods treat the forgotten concept as isolated. In this paper, we study what happens to the rest of the model when a class is forgotten, using a label-conditioned energy-based model (EBM) that assigns per-class energies, making the effect directly observable. We forget a class by raising the energy of its image-label pairs, training with a forget term, a retain anchor to the pretrained model, a global margin, and an energy regularizer that stops the energy magnitudes from growing without limit. A propagation term applies the same forget signal to retain samples, weighted by each sample's DINOv2 similarity to the forget class, so forgetting reaches images that resemble it and leaves the rest untouched. We evaluate on two benchmark datasets: 1) On a subset of DomainNet across four visual domains, we forget tiger, lion, and scissors one at a time. Forgetting a class in the sketch domain also erases it from real, clipart, and painting, with forgetting error reaching 98% and 99% for lion and scissors, and the effect carrying over to the most similar class. 2) On CIFAR-10, we turn off the propagation term and forget each of the ten classes on its own. Forgetting is complete (100%), while the other nine classes retain 98.5% of their pre-unlearning accuracy on average.

Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models but relies on costly rollout exploration. Assigning the same exploration budget to samples with different difficulty levels is inefficient: easy samples may receive redundant rollouts, whereas difficult but learnable samples may receive too little exploration. Existing adaptive schedulers address this mismatch through curriculum-based sample selection or non-uniform rollout allocation based on estimated sample difficulty. However, obtaining reliable online difficulty estimates remains challenging: dedicated probing adds substantial generation overhead, whereas history-based estimators face a cold start with no initial observations and stale feedback, and typically ignore relations among samples. To address these limitations, we propose a plug-and-play graph-based online difficulty estimator that shares rollout feedback across related samples and continuously updates their difficulty estimates, mitigating cold start and staleness without dedicated probing. Specifically, we first construct a difficulty-aware sample graph based on semantic and reasoning similarities. Based on this graph, we introduce latent difficulty states and use a Potts prior to encourage neighboring samples to share the same state. We then employ a state-level Beta-Binomial model to aggregate the rollout outcomes associated with each state. Finally, we use an online mean-field variational algorithm to continuously update the latent-state assignments and state-level difficulty as new feedback arrives. Our framework can be integrated into sample-selection and rollout-allocation schedulers, enabling difficulty-adaptive exploration without dedicated probing. Experiments across multiple base models, RL schedulers, and benchmarks demonstrate that our framework achieves better performance.

Surgical performance assessment in minimally invasive surgery largely relies on manual expert review, making it time-consuming, subjective, and difficult to scale. While existing surgical video understanding methods address tasks such as instrument segmentation, surgical phase recognition, and action recognition, they do not explicitly capture fine-grained tissue handling, a key indicator of surgical quality. To address this gap, we introduce tissue tension recognition, a new clinically motivated video understanding task for laparoscopic and robot-assisted rectal cancer surgery. To support this task, we construct SurgTension, the first expert-annotated tissue tension dataset, providing a benchmark for objective tissue tension recognition. We further propose TensionTRAC, a lightweight trajectory-based framework that models tissue tension from sparse point trajectories. Using a compact trajectory encoder, TensionTRAC achieves competitive performance against strong pretrained video backbones.

Financial time series exhibit non-stationary and heterogeneous statistical properties, making change-point detection challenging because no single unsupervised algorithm performs consistently across assets and market regimes. Conventional workflows consequently depend heavily on expert-driven model selection, feature design, and hyperparameter tuning, limiting their scalability and adaptability. We propose EvoTS-Agent, a validation-guided self-evolving LLM agent for autonomous financial time-series change-point detection. EvoTS-Agent first performs curated exploratory data analysis to characterize dataset properties and initialize candidate detection models. It then evolves executable experiment trajectories through three complementary operators: \textit{Revision} exploits the current best solution, \textit{Alternative Strategy} explores fundamentally different modeling directions when progress stagnates, and \textit{Recombination} synthesizes complementary evidence from high-performing trajectories. Validation feedback guides trajectory evolution throughout the search, enabling the agent to adapt its detection pipeline to the statistical characteristics of each dataset while preserving reliable optimization. Experiments across four benchmark datasets demonstrate that EvoTS-Agent consistently outperforms existing LLM-based agents while maintaining a 100\% execution success rate across all evaluated backbone LLMs.

Groups routinely complete projects that no single member can plan, execute, or verify alone. We propose a formal model of this phenomenon, Collective Counterfactual Planning (CCP), in which the binding limitation on each agent is neither capability, knowledge, nor observability, but representational geometry: each agent perceives the state, conceives moves, consents to actions, and certifies goal requirements only through a projection onto an agent-specific subspace of a common task space. Four gates jointly determine whether a team can reach a conjunctive goal and legitimately recognize that it has done so: the exogenous implementation coalitions required to perform each action, together with three representational gates -- conception, consent, and task-relative verification qualification. We define the Collective Counterfactual Solvability (CCS) problem, separating geometric feasibility, executable attainment, and validated completion. The results expose a positive-negative duality. Iterated cross-agent relay can unlock a solution that no one-shot pooling of individual plans contains, but any goal requirement depending essentially on the subspace dark to the entire team is unverifiable and therefore not validly completable, even when the trajectory accidentally attains it. Memoryless and audited consent further constrain different objects -- action directions versus cumulative trajectory states -- and neither dominates the other. A four-step exhaustive horizon-bounded solvability scheme is sound and complete under exact representation of the relay closure; restricted implementations remain sound on returned plans but need not be complete. The model gives one geometry for sequential mutual enabling, competent execution of steps whose purpose is invisible to the executor, forced sub-teaming at expertise boundaries, and completion that cannot be validly declared.

Recent advances in AI have revolutionized speech processing, yet effective speech understanding requires discerning not just what is said, but how it is said. Speech Sentiment Analysis plays a critical role in decoding these paralinguistic cues for diverse real-world applications such as recruitment and customer service. However, existing Speech Sentiment Analysis research faces two primary limitations. First, dominant approaches rely on text-centric pipelines that cascade Automatic Speech Recognition with text analysis. This process inevitably discards essential acoustic features like prosody and tone, failing to capture attitudinal meanings in acoustically ambiguous utterances. Second, current benchmarks suffer from a mismatch in label granularity, prioritizing basic emotions (e.g., happy, sad) over the nuanced interpersonal stances (e.g., confident, impatient) necessary for social sensitivity. To address these limitations, we propose a novel dataset, SpeechSense, for fine-grained speech sentiment analysis. Specifically, we define a specialized 8-class taxonomy of interpersonal stances detectable primarily through prosodic cues beyond lexical content alone. We then construct a curated dataset based on this taxonomy, built from high-fidelity speech synthesis and rigorous human validation. Comprehensive experiments across multi-modal LLMs, text-only LLMs, and speech encoders demonstrate that models with acoustic access consistently outperform text-only baselines. These results empirically validate the primacy of acoustic cues in detecting subtle speaker attitudes, highlighting the necessity of SpeechSense. Dataset and supplementary materials are available at https://github.com/Sher13cked/SpeechSense.

In the Lifelong Multi-Agent Path Finding (L-MAPF) problem, agents must repeatedly move from one destination to another while avoiding obstacles and inter-agent collisions. Widely regarded as one of the highest-performing solutions to this problem is the Rolling-Horizon Collision Resolution (RHCR) framework. However, commensurate with its quality solutions, it incurs a computational cost that limits its applicability to even modest agent counts. In this paper, leveraging theoretical methods from the Locally Interdependent Multi-Agent MDP literature, we first theoretically prove the near-optimality of RHCR in a discounted MDP formulation of the L-MAPF problem. Then, we leverage these results to naturally motivate an extended framework called Group Decentralized RHCR (GD-RHCR) which incorporates a group decentralized structure that partitions agents based on a transitive communication scheme and plans for each partition of agents in parallel. We show that both RHCR and GD-RHCR achieve similar exponentially close to optimal guarantees, establishing a theoretical duality between the time based restrictions performed by vanilla RHCR and the additional space based partitioning performed by GD-RHCR. Lastly, we show that across varying maps, GD-RHCR is able to attain high throughput that scales into higher agent counts while maintaining a significantly lower per plan cost.

Automatic Target Detection and Recognition (ATD/R) is critical for military decision support and (semi-)autonomous operations. Recent advances in object detection and artificial intelligence (AI) significantly boosted the potential performance of ATD/R. However, the scarcity of publicly available military datasets limits the application of these systems. As a solution, this paper explores the use of publicly available models and civilian datasets to achieve reasonable performance in military contexts. We benchmark several state-of-the-art models, including six iterations of the YOLO series and two variations on the DETR framework, on a newly acquired military relevant dataset. This dataset features military vehicles and challenging circumstances, including various degrees of occlusions and small targets. The out-of-the-box version of each model is validated alongside a version finetuned on the VisDrone dataset. This dataset features small objects, an Air-to-Ground (A2G) perspective and relevant classes, potentially generalizing to our military ATD/R task. We compare the performance of the models using mAP@0.5 and mAP@0.5:0.95, across A2G and Ground-to-Ground (G2G) perspective, target size and model size, giving insight into the real-time capabilities of models. Our main findings are: (1) bigger models outperform smaller models, (2) DETR-based models show promising results compared to the YOLO series,(3) fine-tuning models on an out-of-domain A2G dataset, improves their A2G performance and slightly improves their performance on small objects, but (4) all models still struggle with detecting small objects in an A2G scenario. We conclude that, despite recent advances in object detection, in-domain training is still crucial for creating capable ATD/R systems.

As LLM agents operate across structured workflows and sessions, preserving long-term history does not ensure that later contexts can recover relevant evidence through a bounded memory interface. We study this evidence-reachability problem in long-term conversational memory, where retrieval still relies heavily on semantic similarity. This works well for topical recall, but it often misses earlier experiences, plans, or motivations that are semantically distant from the later events they help explain. Existing memory graphs provide cross-memory structure, yet links driven mainly by semantic overlap can duplicate what the host retriever already recovers. We argue that link construction should instead prioritize a sparse set of retriever-complementary associations. We present CABLE (Complementary Antecedent-Based Linking and Expansion), a plug-in augmentation that constructs links designed to extend the host retriever's direct semantic reach. For each new memory, CABLE generates antecedent-oriented queries, retrieves prior memories, subtracts candidates in the direct semantic neighborhood, and verifies the remainder before adding the accepted complementary associations into a sparse directed graph. At retrieval time, CABLE expands the host system's retrieved seeds along these links to surface implicit supporting evidence. We evaluate CABLE with A-MEM on LoCoMo and MA-LongMemEval, and further integrate it into SimpleMem and Mem0g on LoCoMo, using Qwen3.5-27B, DeepSeek-chat, and GPT-4o-mini. CABLE yields higher mean LLM-judge scores in every evaluated system-level setting, with the largest gains in categories where useful evidence is distributed across memories or sessions, including open-domain, multi-session, and preference-oriented questions. These results support prioritizing sparse, reasoning-relevant associations that complement rather than duplicate the host retriever.

Autonomous research systems are increasingly capable of executing long research workflows, yet automation alone does not ensure that the resulting process remains scientifically grounded. We introduce AutoResearch, a two-stage system that connects Idea Generation with Idea Execution to address both how research ideas are formed and how they are reliably established through experimentation. In Idea Generation, AutoResearch continuously integrates emerging research signals with accumulated domain knowledge, identifies transferable mechanistic insights, and uses multi-model generation and cross-review to produce grounded, testable research plans. In Idea Execution, coordinated agents decompose these plans into experiments, iteratively implement and diagnose them, and employ independent evidence-based review before accepting research conclusions. Across representative settings in cross-modal retrieval, systems optimization, and benchmark-driven machine learning, AutoResearch turns generated ideas into measurable progress, detects and corrects unreliable experimental results, and makes evidence-conditioned decisions to continue, revise, or terminate research directions. For example, on RSICD benchmark, an AutoResearch-generated idea improves mean Recall from 32.84 to 34.69, while recording only 5 audit-confirmed issue events compared with 11-27 for other autonomous research systems. These results demonstrate a research process in which meaningful insight is grounded before experimentation and conclusions are grounded before acceptance: Insight In, Hallucination Out.

Recurrent models process long contexts efficiently by compressing their history into a fixed-size state, but modern architectures typically do so in a single causal pass over the sequence. Each input must therefore be compressed before the model knows how it will later be used, forcing a limited state to compromise across possible future demands. We introduce dynamic compression, which allows a recurrent model to selectively revisit past tokens and revise its fixed-size state through additional recurrent updates. The model need not preserve every part of the history at uniformly high fidelity in its recurrent state, because lower-fidelity information can be revisited from the retained raw sequence when it becomes relevant. We study this in a controlled setting where the model first learns multiple functions in-context and, later in the same sequence, encounters a series of few-shot tasks that each require it to identify and reuse one of those functions. A single-pass model must preserve every function at sufficient fidelity for any future task, whereas selective re-scanning allows the model to revisit and refine only the function currently needed. We find that dynamic compression substantially reduces the recurrent state required for accurate reuse and scales more favorably as the number of stored functions grows. These results demonstrate a computation--memory tradeoff in which recurrent models can spend more computation revisiting their history to make more effective use of a fixed-size state.

While Multimodal Large Language Models (MLLMs) have made significant strides in visual comprehension, their ability to reason about text-dense, professional documents remains incompletely evaluated. Existing benchmarks emphasize information extraction, require external domain knowledge, or cover professional documents only as one of many settings. They are also largely English- or Chinese-centric, leaving other languages and Russian, in particular, substantially underrepresented. To address these limitations, we introduce BEAR-Bench (Bilingual Enterprise and Academic Reasoning), a self-contained, complex English-and-Russian benchmark comprising 1000 human-annotated questions based on text-rich business and scientific documents. We evaluate 16 proprietary and open-weight MLLMs, including Gemini 3.1 Pro and Qwen3.5-397B, on BEAR-Bench and observe clear headroom even for the strongest systems. Finally, we use the resulting model outputs to compare existing hallucination detection methods, evaluating not only how often models fail on BEAR-Bench but also how reliably those failures can be identified.

This paper investigates the feasibility of using residual learning to improve unsteady aerodynamic load prediction for aeroelastic applications. The machine learning technique selected for the study is the long short-term memory (LSTM) neural network, which is used for its suitability for sequential data with aerodynamic memory effects. The approach is investigated for the NLR 7301 airfoil benchmark using high-fidelity CFD lift data for prescribed pitch and plunge motions in the transonic flow regime in the presence of shock motion. An analytical unsteady aerodynamic model based on the Wagner function is used as a physics-based baseline, and the neural network is trained to learn the difference between the CFD lift coefficient and the Wagner prediction. The residual model is compared with a direct neural-network model trained to predict the CFD lift coefficient. The comparison includes feature and normalization studies, external benchmark cases, and leave-one-out and leave-family-out generalization tests across a range of sinusoidal and non-sinusoidal motions. The residual model performs best when its inputs align with the Wagner formulation variables, generally giving lower error and more consistent performance across training runs, though the direct model remains more accurate for some high-frequency cases. The residual model also generalizes better in the leave-one-out and leave-family-out tests, with a smaller increase in error than the direct model when entire motion families are withheld from training. Overall, the results indicate that residual learning shows promise as a modular approach for augmenting classical low-order aerodynamic theories, especially when the physics baseline removes a structured part of the aerodynamic response and leaves a lower-variance correction for the neural network to learn.

Trajectory prediction is central to safety in autonomous driving, yet learning-based predictors tend to degrade sharply when encountering scenarios poorly represented by their training data. Many methods attempt to mitigate distribution shift degradation through data-centric or test-time adaptation approaches; however, they are typically validated along fragmented axes of generalization, leaving the field without a standardized way to compare robustness across shifts a model may encounter. To address this, we introduce ControlledShifts, a framework and benchmark suite that systematically re-splits existing trajectory datasets into in-distribution (seen) and out-of-distribution (unseen) partitions, via a shared characterization-and-splitting formulation, in which a characterization function fixes the axis of variation a benchmark probes and a splitting function fixes how the tail of that axis is withheld. The suite comprises three benchmarks targeting key topological and behavioral distribution shifts. Furthermore, to aggregate multi-dimensional performance metrics across these benchmarks, we propose a unified robustness score that evaluates models along two complementary dimensions: prediction quality (relative performance gain) and prediction stability (performance preservation under shift). We showcase ControlledShifts by benchmarking prominent transformer-based architectures, exposing critical differences in how models of varying capacities handle latent relevance and environmental structure.

Many high-performing pathology tile encoders are now foundation models with hundreds of millions to over a billion parameters. Encoding and storing the thousands of tiles in each whole-slide image with such models is costly on commodity hardware, so compact encoders that retain useful downstream performance are a valuable alternative. We present DistillPath-KS16, which starts from the existing 22M kaiko ViT-S/16 encoder and improves it by distilling from released pathology encoders used as frozen teachers. The recipe reads only the teachers' final class and patch tokens and trains on 6,000 public slides, needing neither their DINO nor iBOT pretraining heads nor a billion-tile corpus, so it applies to any released encoder that exposes backbone tokens. We distill four teachers spanning 86M to 1.1B parameters into the same student. Every variant improves the kaiko baseline on all three benchmarks we use, EVA, HEST, and PLISM, and the strongest teacher is task-dependent. On the seven-task EVA mean, DistillPath-KS16-Virchow2 reaches $0.795$, within $0.015$ points of Virchow2, the top-scoring model in our evaluation, at about $29\times$ fewer parameters; it also scores above H0-mini and GPFM on this aggregate metric, though that advantage is task-concentrated rather than uniform. Because it remains a 22M ViT-S/16 with 384-dimensional features, DistillPath-KS16 runs more than $25\times$ faster than Virchow2. Code is available at https://github.com/RamonKaspar/DistillPath, and released model weights are available at https://huggingface.co/collections/RamonK/distillpath.

Tabular prediction is a critical task across numerous applications. The recent success of large language models has sparked various approaches for adapting them to the tabular domain. A prevalent strategy involves training or fine-tuning specialized Tabular Foundation Models (TFMs) such as TabPFN. However, TFMs require substantial computational resources, and frequent model retraining is often impractical. In-context learning (ICL), specifically, few-shot prompting, offers a resource-efficient alternative to enhance performance. Yet, identifying the most relevant rows to serve as shots remains a challenge for tabular data. This paper introduces ARASH (Adaptive, query-specific Retrieval And Shot selection), a method that improves TFM efficiency by selecting optimal shots based on local neighborhood analysis within the training set. Our results demonstrate that ARASH reduces the prompt length and memory usage of TabPFN by 1261.5$\times$ and 2.56$\times$, respectively, while providing comparable accuracy.

Large language models are pushing data science toward increasingly autonomous and agentic workflows, with recent systems already supporting multi-step and long-running analyses. As these workflows become more autonomous, conventional interfaces no longer provide adequate support for two critical requirements: observability for understanding an agent's evolving reasoning and evidence, and steerability for redirecting low-value directions or deepening promising ones during execution. Existing interactive approaches improve process visibility and open intervention points, but they remain largely designed for discrete, turn-by-turn exchanges rather than the parallel branches and evolving decision structures of long-running agentic analysis. We study this need as interactive oversight in long-running agentic data analysis and present AdaLens, an interactive system for monitoring and steering ongoing runs. AdaLens combines a storyline-based representation that unifies analytical plans, execution progress, intermediate findings, and data-column involvement with steering interactions grounded in these analytical elements for directional guidance and execution control. We evaluate AdaLens through two case studies and a user study, examining how it supports analysts in monitoring and steering long-running agentic data analysis.

Public institutions face a persistent challenge in selecting LLMs suited to their specific context. Existing benchmarks, however, are of limited use as they primarily reflect English-language and US-centric settings, and often only evaluate task performance. In this paper, we present first results of MÖVE, a holistic evaluation framework for the German public sector, examining three rarely considered governance dimensions: energy consumption, provider transparency, and knowledge of German-party positions. Our results reveal significant trade-offs, with no single model excelling across all dimensions: estimated energy consumption varies more than 60-fold and is not explained by model size alone, information disclosure varies systematically across providers, and European models do not exhibit stronger knowledge of German party positions. Model selection for public institutions thus cannot rely on performance rankings alone. Instead, evaluations should also reflect the governance requirements of the deployment context.

Aerial manipulation extends robotic operations to previously inaccessible aerial environments. Unlike arm-equipped aerial systems, tiltable-multirotors can directly generate six-degree-of-freedom wrenches through their flight bases, enabling both efficient movement and omnidirectional operation by tilting the thrust direction. This work presents a design analysis and a wrench-based control framework for tiltable-multirotors in aerial manipulation. We show that a four-rotor tiltable configuration provides a balance between interference-free propeller sizing and hovering efficiency across different attitudes, and its null-space redundancy is crucial for traversing singular configurations under physical constraints. We further show that an upward end-effector placement yields a favorable trade-off between geometric clearance and available wrench. To address disturbances, we propose a dual strategy consisting of a modified integral term for model error and an acceleration-based estimator for external wrenches. Building on these insights, we develop an effector-centric nonlinear model predictive control (NMPC) framework that integrates design choices, singularity handling, and disturbance compensation into a unified formulation. The proposed framework runs fully onboard at 100 Hz on a custom-built tiltable-quadrotor. Real-world experiments, including a 90-deg step cartwheel rotation, whiteboard pushing, and continuous 360-deg valve turning, demonstrate the feasibility of wrench-based omnidirectional manipulation with singularity traversal on a one-DoF-per-arm tiltable-quadrotor.

Reliable transfer of orbital angular momentum (OAM) to microscopic objects typically relies on high-fidelity vortex beams generated by programmable spatial light modulators or precision-fabricated phase optics. Here, we demonstrate that robust OAM transfer in optical tweezers can be achieved using static binary holograms printed on acetate substrates. The printed diffractive optics generate Laguerre-Gaussian vortex beams with sufficient spatial fidelity to induce controlled optical torque and stable rotational manipulation of polystyrene microspheres in a high-numerical-aperture optical tweezers system. Despite a diffraction efficiency of only approximately 2%, the generated beams enable reproducible particle rotation using less than 1 mW of optical power in the first diffraction order. The rotational dynamics were systematically characterized as a function of incident optical power and topological charge, revealing the expected increase in angular velocity with both parameters, consistent with OAM-driven torque in the overdamped regime. These results demonstrate that efficient optical angular momentum transfer is remarkably tolerant to the reduced efficiency of passive printed diffractive optics, establishing a robust, scalable, and high-damage-threshold platform for structured-light optical manipulation with applications in microfluidics, biophysics, optomechanics, and optical trapping.

Practical LLM unlearning is usually evaluated through two objectives: suppress target-specific knowledge and preserve non-target utility. In generative QA, this leaves a third behavior underspecified: when a target-adjacent prompt admits a broader answer without target-specific leakage, the model should answer at that level rather than leak, evade, or refuse. We study this specification problem in a controlled LoRA-GRPO RWKU setting, comparing four reward designs that span lexical suppression, anti-refusal shaping, rubric-based broad answering, and an explicit refusal contrast, with and without SFT warm-up. The experiments show that optimization success is not equivalent to behavioral unlearning: RWKU forget scores, held-out completion audits, terminal training-rollout audits, and training dynamics can point to different conclusions. We trace these disagreements to reward-hacking endpoints, policy-support limits in GRPO, benchmark probes that miss endpoint changes, and rewards that can select broad-topic answering with low semantic leakage during optimization.

Recent advances in Large Language Models(LLMs) and agents have substantially improved the ability of AI systems to execute complex tasks. Yet existing benchmarks largely rely on researcher-selected tasks, leaving uncertain whether such progress extends to the work that real-world users actually demand from AI systems. We introduce \textbf{StartupBench}, an E2E agent benchmark grounded in market-validated AI startup products. Rather than defining tasks from pre-defined assumptions about useful agent capabilities, we systematically study AI products with demonstrated adoption, together with their product workflows and users, to identify real-world tasks for which AI has established practical demand across diverse professional domains. We translate these workflows into complete deliverable-oriented tasks and evaluate them with fine-grained rubrics capturing their complex requirements. Across representative models evaluated under a unified agent harness, even the strongest model successfully completes only approximately 30\% of StartupBench, despite making substantial partial progress on many tasks. Further analysis identifies aspects like complex instruction following and domain-specific expertise as major sources of failure. Our results reveal that many market-validated workflows remain beyond the reliable capabilities of current general-purpose agents, establishing StartupBench as an empirical measure of progress toward E2E completions of real-world user tasks.

Dynamic vision sensors (DVS) are enticing candidates to reach the low-latency, sub-ms target of edge-vision applications, as they generate events with a $μ$s-level time resolution. However, using DVS front ends also calls for novel algorithm/hardware back ends capable of efficiently handling streams of sparse spatiotemporal events. While event-driven graph neural networks (EV-GNNs) have emerged as a solution on the algorithmic side that is both accurate and efficient, there is no dedicated hardware to date capable of efficiently supporting their mixed requirements of dense-regular compute operations and sparse-irregular memory accesses. We therefore introduce ETHEREAL, the first EV-GNN processor chip, capable of bridging this gap by means of a neighbor-parallel spline-convolution engine combined with a split-2D/3D memory hierarchy that introduces a novel spatiotemporal event-caching mechanism. Measurement results demonstrate a 25.6$μ$s latency and a 1.6$μ$J energy per end-to-end event-wise inference on the state-of-the art DAGr-GNN workload and VGA-resolution (640x480 pixels) DSEC dataset.

ML systems increasingly condition decisions on downstream model identity, but this is useful only if model-specific differences form reusable structure rather than input-local interactions. We test this in retrieval-augmented generation (RAG), where evidence utility can be measured under controlled interventions. Holding query, evidence, task, scoring, and intervention fixed, nine readers disagree on effect sign in 33\% of jointly affected cells; reader$\times$query interaction explains 29.8\% of utility variance versus an 8.4\% permutation null; and self-selected evidence improves F1 by $+0.031$ ($t=3.39$). We then ask the sharper question: \emph{which components of this heterogeneity are stable reader properties across queries?} Separating three measurable objects---evidence \emph{activity}, \emph{ordinal preference}, and \emph{conditional signed direction}---we find ordinal reader geometry stable across four independent settings (split-half $ρ=0.60$--$0.83$): leave-one-out interventions, PRISM preferences, RAMDocs, and RAGuard. Signed geometry is task-bounded: weak in open-ended QA (0.14, 0.35), especially for misleading and irrelevant evidence, but strong in binary fact-checking (0.75) with no significant ordinal gap, though still below its sparsity-matched ceiling. Sparsity, decoding noise, and metric artifacts do not explain the main ordinal--signed gap. Finally, stable ordinal similarity fails to predict cross-reader intervention transfer (oracle-distance $ρ=-0.27$; regret reliability $-0.28$). Reader-specific utility exists, but preference is not intervention: stable ranking similarity does not license transfer of help/harm decisions.

We demonstrate that RL finetuning an LLM using debate, a two-player adversarial game between a generator and a critic adjudicated by a weaker LLM judge, reduces reward hacking compared to a reinforcement learning from AI feedback (RLAIF) baseline. Reward hacking is a central obstacle in RLAIF: as training progresses, the policy learns to exploit systematic errors in its AI judge, degrading task performance, a problem that worsens precisely when the judge is weaker than the policy, the setting most relevant to overseeing increasingly capable AI systems. We study mathematics tasks, where final-answer correctness is verifiable, allowing us to measure reward hacking dynamics. We train a Gemini~2.5 Flash-class policy with a frozen, weaker Gemini~2.5 Flash Lite judge, comparing a single-player RLAIF baseline against debate. While the baseline quickly hacks the judge, debate maintains judge performance throughout training, leading to a higher peak validation accuracy (45\% performance gap recovered) that persists through many RL steps. Additional experiments show that: 1) further weakening the judge leads to faster hacking, but this can be compensated by adding an additional debate round; 2) debate incentives override prompted misalignment; 3) RL using an LLM judge has a smaller train/validation reward gap than RL from verifiable rewards; 4) learning to critique to convince the judge using ground truth labels is possible but slow. Taken together, our results are a positive update on the feasibility of debate, while highlighting that balancing multi-agent training is critical: without player constraints, adversarial training risks defaulting to critic judge-hacking. We show that critique word limits (effective up to 150 words) successfully balance the game and avoid judge hacking, though this introduces a trade-off by restricting critic expressive clarity.

Anomaly detectors are hardest to deploy exactly where training data is scarcest: a newly commissioned production line has a handful of verified "golden" samples and no machine-learning engineer on the factory floor. We present a training-free human-in-the-loop framework in which a domain expert corrects a PatchCore detector by direct memory bank editing: no retraining, no gradients, no original training data. A false-positive correction inserts the reviewed image's normal patches through a self-calibrating novelty gate admitting only those beyond the median pool-normal nearest-neighbour distance. From a bank built on only ten golden samples, operator corrections close a median 66% of the gap to an uncorrected fully trained bank (mean 80%, raised by three categories that overshoot parity), significantly improving 12 of 15 MVTec AD categories and harming none: ten samples plus corrections outperform hundreds of samples without them. On already-trained banks the headroom is smaller and concentrated where the bank undersamples normal appearance (gated: toothbrush +0.10, metal nut +0.09, zipper +0.05, screw +0.05), and no category except grid is significantly harmed. Evaluation uses a held-out protocol (20 splits per category, Holm-corrected Wilcoxon), because corrected images entering the bank inflate naive evaluation toward AUROC 1.0 by memorisation. Passive and active querying are statistically indistinguishable; a matched-label-budget control attributes gains to deployment-time label production at 43% of exhaustive-review cost; a defect-memory extension fails decisively. Feedback is simulated from ground truth; live expert trials, where mislabelling is costliest on small banks, remain future work.

Embodied intelligence is shifting artificial intelligence from passive digital perception toward active physical interaction. However, foundation-model-enabled embodied agents face a fundamental tension between open-world cognition and resource-constrained deployment. On-device models are limited by computation, memory, and energy budgets, whereas cloud-centric solutions introduce latency and reliability risks over dynamic wireless links. Edge general intelligence provides a promising cognitive backbone, but existing frameworks still lack physical grounding, action awareness, and mechanisms for actively acquiring useful physical experience. To address these limitations, this article introduces edge-native embodied intelligence (ENEI), an action-aware wireless edge framework that integrates embodied agents, the 6G communication and networking fabric, and edge cognitive services into a 6G-mediated bidirectional edge-embodiment loop. Along the edge-to-embodiment axis, confidence-aware assistance and edge-driven generative adaptation enhance local autonomy under out-of-distribution (OOD) conditions. Along the embodiment-to-edge axis, value-of-experience guided active embodied federated learning enables physical actions to generate informative experience for continuous edge model evolution. The 6G fabric supports both directions through goal-oriented transmission and programmable radio-resource allocation. Two case studies on OOD drone navigation and mobility-driven federated learning illustrate the feasibility and communication efficiency of the proposed mechanisms. ENEI provides a unified perspective in which edge cognition strengthens embodied action, while embodied agency actively enriches edge cognition, laying the foundation for scalable, adaptive, and self-evolving embodied wireless systems.

Memory is a key capability of LLM agents. Persistent memory extends this across sessions---enabling recall, revision, and personalization. Yet its multi-stage pipeline (ingestion, retrieval, filtering, generation) makes failures difficult to localize: end-to-end evaluation reveals that an error occurred, but not which stage caused it. Existing evaluations often report aggregate performance without paired statistical comparisons, slice-level non-regression checks, or stage-level diagnostic traces. We propose D$^2$ACCI (Diagnostic-Driven Artifact-based Closed-loop Controlled Iteration), a dual-loop protocol whose outer diagnostic gate promotes, feature-flags, or rejects memory interventions based on paired evidence, protected-slice monitoring, and trace-level localizability. We further introduce DCR, a graded observability metric that measures whether failures remain localizable, and D$^2$ACCI-Eval, a reusable artifact for gate replay. We instantiate the protocol in MemStack and evaluate on three public benchmarks, achieving 93.59% on LoCoMo, 90.93% on LongMemEval, and 57.20% on PersonaMem-V2. Five paired ablations show that supplement extraction, session-memory retrieval, and Forget Guard yield statistically significant gains (+1.9 to +3.7pp, all p $\le$ .003). In contrast, BM25/RRF is retained as a monitored feature flag---a distinction invisible to aggregate-only evaluation. A diagnostic audit shows enriched traces substantially improve root-cause agreement over result-only relabeling. Diagnostic artifacts reach 98--100% DCR@3 versus 0% for results-only logs. These results establish that robust memory-system iteration demands traceable, statistically grounded, and regression-aware evidence---exactly the gap D$^2$ACCI fills.

Decentralised partially observable Markov decision processes (DecPOMDPs) provide a general framework for modelling multi-agent decision making under uncertainty. However, DecPOMDPs are known to suffer from exponential complexity in the number of agents. One way to combat this intractability in agent numbers is to look at partitions of agents that exhibit a form of symmetry among agents, allowing for a compact encoding by counting. However, a challenge arises as the policy space explodes, even though the model complexity and evaluation cost reduce to a polynomial dependence. In this paper, we redirect our focus from counting agents to counting policies, which actually enables tractability in agent numbers for so called policy-counted DecPOMDPs. Further, we present policy-counted dynamic programming using the compact representation to solve policy-counted DecPOMDPs efficiently.

Take three frontier mixture-of-experts models (Alibaba, OpenAI, NVIDIA; 3.6-4.0B active parameters each) and fine-tune them to reason in a low-resource language. On accuracy benchmarks almost nothing happens, and the benchmark itself is noise at this scale: changing only the random seed moves the score by 7.7 points, more than every data and recipe effect we measured. That null is our first result. The real changes live where accuracy cannot see. Base models never think in Greek: 0 of 1,000 reasoning traces, even when the question is Greek, so the model answers correctly while reasoning in a form its user cannot read, audit, or correct. After supervised fine-tuning (SFT), every released checkpoint reasons in the language of the question on ~98% of items, one family at 3x fewer tokens, with judged grammaticality improving on all four models and general ability within a few points of each base: nothing was forgotten, and fluency was gained. We propose six behavioural dimensions that make such changes measurable, each gated to reject any metric that correlates with output length, and we report how our own instruments lied: six failures, each caught by a control. What SFT cannot do is fix its own defects: a quarter of answers skip the requested format, answers leak into the reasoning channel, and an explicit "think in English" is obeyed under half the time. Reinforcement learning with verifiable rewards, pre-registered before training, fixes the first two outright (fallback 24% to 2.5%, leak 3.5% to 0.0%, both against a flat random-reward control) and moves the third (+9.1pp), while the Greek reasoning habit survives an accuracy-only gradient untouched. We release five checkpoints. The instruments, the controls and the pre-registration travel to any low-resource language; Greek is the case that let us measure them.

This study investigates cooperative control of connected and automated vehicles (CAVs) at partially observable highway bottlenecks in mixed traffic, aiming to mitigate congestion without relying on complete global traffic states or online trial-and-error. We propose a physics-informed world model-based offline multi-agent reinforcement learning framework that reconstructs a physically interpretable global traffic state from local CAV observation-action histories, with coupled macroscopic-microscopic traffic dynamics providing physics-based supervision. A probabilistic ensemble world model learns traffic-state transitions and system rewards, while model disagreement quantifies epistemic uncertainty. Multi-step imagined rollouts with pessimistic rewards and uncertainty-driven truncation are then used for offline policy learning. Experiments in a SUMO-based on-ramp bottleneck using approximately $1\times10^6$ offline transitions show that physics supervision improves state reconstruction and world-model prediction accuracy.

In bus routing, the task is to plan a bus route in a network with several agents, each of whom wants to travel from a starting point to a destination. A bus route should account for several factors, including agents' cost for reaching the bus stops, their travel time, or the energy consumption of the buses. We study the complexity of several variants of this problem, focusing on how the objective function and the models for agents' walking costs influence the problem complexity. After observing that even the simplest agent cost model leads to hardness on general networks, we consider networks with tree structure. Our main findings are as follows. First, allowing agent-specific cost models leads to hardness even on extremely limited trees such as stars. Second, consistent agent models (where agents differ only in their starting points and destinations) make the problem easier in some cases. Finally, allowing agents to choose between using the bus and walking directly can make the problem considerably harder. Most of our hardness results show not only classical NP-hardness but also parameterized intractability for the natural parameter $k$, the number of bus stops.

Bridge structures are regularly inspected for structural damage such as cracks and corrosion in order to ensure public safety and reduce maintenance costs. Much research has been done on automating this process using computer vision methods, which are often evaluated and compared using metrics such as intersection over union, mean average precision, etc. However, predicting the actual effectiveness of an inspection method within the field of structural engineering from these metrics remains challenging. To enable the systematic use of these increasingly popular methods in engineering practice, evaluating the performance of these methods in a way that is compatible with standard engineering approaches is therefore an urgent necessity. We present a new statistical evaluation framework to allow the comparison of computer vision methods with conventional visual inspection for crack detection in steel bridges. The framework is based on probability of detection curves and can account for the influence of image resolution. We apply this evaluation method to the real-world ``Cracks in Steel Bridges'' dataset, which contains annotated images of cracks in bridge structures. The quantification of the probability of detection and its uncertainty enables a practical assessment of the effect of automated methods for damage detection in structural reliability analyses. In turn, this enables the wide-spread use of automated (AI-based) damage detection in safety critical applications. This evaluation method provides evidence that the proposed computer vision approach approach is robust for the crack detection task and can have a high added value as an addition to conventional visual inspection methods.

Context: Software systems that depend on commercial large language model APIs must migrate to successor versions when vendors deprecate older models. Migration decisions typically rely on aggregate benchmark scores, which compress heterogeneous item-level behaviour into a single net figure. Objective: We measure what that compression conceals. Method: On three pairwise upgrades in the GPT-5.4 to GPT-5.6 Sol product sequence, we query 900 public benchmark items (graduate-level knowledge, olympiad mathematics, instruction following) 50 times per item per model, classify each item as reliably improved, reliably regressed, practically equivalent, or inconclusive under false-discovery-rate control and a practical-significance threshold, and calibrate the results against a label-permutation null. Results: Across all nine migration-benchmark cells, reliable improvements and reliable regressions coexist. Edges with aggregate gains of up to 7.3 percentage points contain up to 8.3% reliably regressed items; edges with aggregate losses contain up to 10.7% reliably improved items. On the instruction-following benchmark, the gap between strict and loose scoring widens by 3.9 percentage points on the latest migration: a 3.9-point regression under strict scoring shrinks to 0.04 points under loose scoring. Conclusion: Migration decisions based on aggregate scores alone miss substantial bidirectional item-level change. The complete response-level archive and per-item scoring outputs are released.

Long-horizon agents increasingly operate across many steps, tools, and observa- tions. In this setting, the relevant oversight question is not only whether each action is locally valid, but whether the evolving trajectory still corresponds to the task the user authorized. Drift can accumulate quietly: an agent may call the right tool with plausible arguments at every step, while its prefix moves toward a broader role, an adjacent objective, or evidence the user never supplied. Existing monitors mostly check local compliance, deliver final-trace verdicts, or score generic risk; they do not directly estimate this prefix-level relation. We introduce ontological trust, a task-conditioned property of trajectory prefixes, and instantiate it as RGE, an online monitor that decomposes trust along Role, Goal, and Evidence. RGE uses LLMs only to derive structured task and step representations; trust-state updates, projec- tions, and intervention decisions are deterministic, so the output is a replayable and auditable trust trajectory rather than a single end-to-end judge verdict. We construct a cross-domain trajectory corpus from OSWorld, FinanceBench, and EICU-AC, covering benign executions, prefix-paired drift, and pseudo-consistency failures. On this corpus, RGE outperforms adapted rule-, judge-, and shield-style baselines on prefix-paired drift detection. With the two larger estimator models, it exceeds 93% Drift F1 on every benchmark while keeping benign coverage at or above 95.8%. Pseudo-consistency is harder: detection depends on whether task completion is externally visible, a structural limit we characterize empirically.

A robot asked to "place the cup near the red plate or the blue plate" may reach the centroid between them and appear geometrically successful, while satisfying neither disjunct of the instruction. This silent semantic failure exposes a structural limitation of language-conditioned robot policies: representations that collapse a disjunctive instruction into a single connected set cannot preserve all feasible modes, and planners that commit to one action degrade under run-time mode uncertainty. We address this limitation with CompCPZ, a sound algebraic layer that language-conditioned learning systems wrap to recover multi-modal disjunctive representation, recursively composing per-primitive constrained polynomial zonotope enclosures along the language parse tree with distribution-free conformal coverage and sub-millisecond runtime. On a closed-loop ManiSkill3 tabletop-manipulation benchmark, CompCPZ outperforms convex set baselines, multi-peak decoders, and a zero-shot vision-language-action model (1,900/1,918 paired wins, p << 10^(-30)); the same compiler also transfers without retuning to planar real-robot trials on a Unitree Go2 quadruped under motion capture. These results suggest that compositional language grounding should be evaluated not only by reaching a decoded target, but by whether the represented feasibility set preserves the connected-component structure of the user's intent.

Agent evaluations and trace-based learning often compare outputs across transformed views through a post-response correspondence treated as neutral preprocessing. We show that this correspondence is a measurement intervention: omitting it can manufacture sensitivity, an over-aggressive map can manufacture invariance, and multiple optimal correspondences can leave mechanism labels and signed learning credit unidentified. We develop a validity theory and audit with three components: two-sided validation of nuisance removal and response preservation, all-optima identification of downstream conclusions, and uncertainty propagation after validity is established. We characterize the linear feasibility boundary for response-preserving nuisance removal, compute sharp ranges over exact-optimum correspondence sets, and give a distribution-free certificate that retains a credit coordinate only when all exact optima agree on its nonzero sign. Across public code and SQL pipelines, two deterministic optimal tracebacks disagree on temporal localization for 55.9% of 1,586 nonzero trajectory pairs; two frozen 800-rollout tool-use audits, including a task-and-seed-disjoint replication, expose exact-optimum reversals of intended turn-level credit, although a clean public quick-start subset shows none. A pre-registered transport gate failed on natural responses; frozen corrected and held-out controls then show that a map calibrated only on benign examples erases every retained harmful response, while two-sided validation selects response-preserving alternatives. Cross-view correspondence must therefore be declared, validated, and propagated into uncertainty before agent evaluation or credit assignment supports a point conclusion.

Monitoring nature restoration at scale is an important but difficult ecological problem. Deep learning methods to analyze satellite image time series (SITS) have been widely used for land surface monitoring. In semi-natural grasslands - the habitat type in focus in this work - restoration outcomes develop gradually, yet satellite observations are influenced by weather, acquisition conditions, and processing artefacts, making it difficult to distinguish genuine restoration signals from unrelated temporal variation. In this work, we examine - to the best of our knowledge, for the first time - whether restoration status can be detected directly from satellite image time series by formulating pasture restoration as a binary deep learning classification problem. We evaluate two common SITS deep learning architectures on different Sentinel-2 image combinations, across 1,397 restored Swedish pastures and find that explicitly modeling intra-year variability and per-pasture normalization increases separability, reaching 0.88 accuracy for the best model. We further investigate our results and perform a targeted bias analysis finding that reliable deployment requires temporally balanced labels and evaluation protocols that explicitly test for year-related confounding. We therefore frame our contribution not as a solved restoration-monitoring system, but as a realistic case study of what works, what fails, and what future studies should control for. Code and models are available at https://github.com/aleksispi/ml-nature-resto.

Mobile robots that operate in side by side with humans and critical facilities must reach their goals at low cost, despite often unknown true traversal costs of the map apriori and imperfect actuation. Planners that solve the underlying stochastic shortest path problem exactly, such as value iteration, require computation that grows with the diameter of the map, whereas Dijkstra's algorithm is fast but is usually considered inexact once transitions are stochastic. This study shows that Dijkstra's algorithm can remain an exact planning engine under a condition that is much weaker than the causality condition often invoked in the literature, namely nonnegativity of a reduced cost defined on the determinized map. Building on this characterization, an online learner DORA (Dijkstra Oracle Reduced-cost Algorithm) is proposed for robot navigation that calls a shortest path oracle a fixed number of times per episode, never estimates a transition kernel, and adds a logarithmic survival weight when the probability of contact with a dynamic obstacle must stay within a budget. In the numerical experiments involving three other benchmarks that cover grid world navigation, directional drilling, and drone surveillance, the learner matches optimistic value iteration that is given the true transition kernel while performing 4.5 to 19.3 times less planner work, reduces contacts during learning by a factor of seventeen relative to determinize and replan, and keeps the contact rate within budgets that span two orders of magnitude. These results indicate that shortest path search supports safe and efficient online navigation and path planning tasks.

Autoencoders are commonly used for unsupervised data-driven fault detection in nonlinear dynamical systems. Despite their widespread success and often favorable performance compared with traditional approaches, most applications rely on heuristic reconstruction of measured data using features learned from nominal training data, without explicit insight into the underlying nonlinear dynamics. This lack of interpretability limits the extension of autoencoder-based fault detection methods to higher levels of fault diagnosis, e.g., fault localization and quantification, and confines their use largely to application-oriented studies. To address this limitation, we propose a strategy for detecting parametric faults in nonlinear stochastic mechanical systems. A mathematical representation of the output data is developed using Koopman operator theory, which motivates their embedding on a manifold and its subsequent approximation with a two-stage autoencoder. Fault detection is formulated within a hypothesis-testing framework, in which new data are tested for consistency with a neighborhood of the manifold identified from nominal observations. The proposed method is validated through Monte Carlo simulations of a toy mechanical system with two types of nonlinearity and applied to two well-known real benchmarks, where it provides favorable fault-detection performance compared with standard autoencoders.

Existing LLM-based approaches to Architecture Decision Record (ADR) generation share a critical and largely unexamined assumption: that input is already reasonably structured. In practice, architectural decisions emerge from informal, noisy meetings where choices are implicit, fragmented, and entangled with off-topic dialogue, precisely the conditions under which single-pass prompting degrades. This paper presents GADR, a multi-agent, self-correcting workflow that extracts architectural decisions from raw meeting transcriptions and generates Nygard-formatted ADR drafts. A feasibility study comprising five real project meeting transcripts, expert review by four senior architects, and evaluation by fifteen students provides initial evidence that the agentic workflow captures most expert-identified decisions and produces drafts participants found clear and useful, outperforming zero-shot and few-shot baselines in stability and structural adherence. The study also addresses the underexplored trade-off of RAG-based enrichment improving ADR depth while simultaneously risking transcript-unfaithful content, raising open questions about traceability in automated architectural documentation that we believe is worth the community's attention.

Key-keyway assembly tasks impose strict geometric constraints and are highly sensitive to grasp pose deviations in uncertain environments. This work presents a force-based offset estimation method for keyed peg-in-hole assembly, embedded within a perception-validation-insertion pipeline. Residual misalignment is estimated directly from wrist force/torque measurements using a local KNN-Gaussian Process hybrid regressor. The framework distinguishes between two contact regimes, hard collision and guided chamfer insertion, and routes inference to a dedicated model for each. Regime classification is achieved via a contact-window duration threshold. KNN combined with a deterministic search using the results of a post-grasp monocular visual validation contributes to an increased accuracy of the regressor model. This approach achieves accurate radial offset estimation in chamfered peg insertion, during a keypoint detection-based pick and place application. Experiments using the integrated force/torque sensor of a collaborative robot arm showed an increase in insertion success rate from 67% to 87% after the pipeline was applied.

Multi-robot patrolling requires a team to visit all areas of an environment at regular intervals, typically minimising idleness. A practical extension, motivated by security and environmental monitoring, is to additionally form a collective ranking of all patrol locations by some measured signal, a generalisation of the best-of-n problem to the many-option, continuous-valued regime. We observe that the patrol graph admits a natural dual interpretation: it is simultaneously the topology that dictates agent movement and a factor graph over which spatial beliefs can be propagated. Exploiting this equivalence, we apply Gaussian Belief Propagation (GBP), a graph-based algorithm, to collective ranking using unary measurement factors at visited nodes and pairwise smoothness factors along patrol edges. We compare GBP against simple and visit-count-weighted averaging across a range of sensor-noise conditions in simulation, and validate the approach on four Leo Rovers tracking a propagating radio signal in an office lobby. GBP outperforms both baselines on ranking accuracy, mean squared error, and time to consensus. We find that as noise increases and the task becomes harder, GBP degrades gracefully in simulation while both averaging methods degrade substantially. Hardware trials reproduce the same performance ordering on a real propagating radio signal, supporting the practical relevance of the simulated results.

Self-evolving agents turn experience into reusable skills, workflows, or memories, but post-evolution accuracy alone does not show whether learned behavior preserves previously correct behavior or security. We audit SkillOpt, Agent Workflow Memory (AWM), and ReasoningBank in simulated e-banking using matched benign acquisition trajectories, sealed evaluation endpoints, execution-grounded checks, and independent state replay. On Qwen 3.7 Flash, SkillOpt raises benign utility from 0.741 to 0.837 while exposure to injected content rises from 0.820 to 0.943. Conditional attack success after exposure falls from 0.605 to 0.562, yet overall attack success rate (ASR) rises from 0.496 to 0.530 and unauthorized financial state changes rise to 0.685. Across three independently evolved lineages, capability, exposure, and unauthorized-state changes increase in all three, whereas ASR increases in only two. ReasoningBank raises utility to 0.859 without increasing aggregate ASR, although unauthorized state changes remain slightly above Static. AWM reveals a separate evaluation hazard: a literal WebArena text-action envelope disrupts tool execution in our native function-calling executor. In a post-hoc sensitivity test, removing only that envelope restores utility from 0.319 to 0.756, while exposure rises from 0.299 to 0.909 and ASR from 0.195 to 0.575. Auditing self-evolving financial agents therefore requires tracking regressions, attack-surface contact, unauthorized financial-state change, and artifact-executor compatibility, not accuracy alone.

Real-time novel view synthesis is dominated by rasterized explicit primitives. These projection-based pipelines provide high throughput but require specialized extensions for non-pinhole effects such as distortion, rolling shutter, and depth of field. Ray-based rendering expresses these effects naturally but is generally assumed too slow for competitive real-time rendering. We analyze the factors governing throughput in differentiable Voronoi ray tracing and identify traversal length, per-cell work, and memory locality as principal determinants. Guided by this, we introduce VoroTracing, which co-designs the scene representation, optimization, and GPU execution to reduce these costs. Compact octahedral appearance textures reduce memory traffic, while surface-concentrated opacity promotes early termination. The fixed-budget representation is optimized without pruning or densification and rendered with a GPU implementation designed for coherent traversal. On Mip-NeRF 360, VoroTracing renders at 623 FPS on an RTX 5090, providing $3.2\times$ the throughput of the fastest prior ray-based method and $2.8\times$ that of 3D Gaussian Splatting, while maintaining competitive reconstruction quality. Our renderer supports fisheye, rolling-shutter, motion-blur, and depth-of-field effects through ray generation and sampling, requiring no specialized rasterization. These results show that real-time throughput can be achieved with the flexibility of ray-based rendering. We release our source code, see https://research.zenseact.com/publications/vorotracing

Automated security patch backporting is critical for mitigating N-day vulnerabilities. Recent tools report success rates above 80% on their respective datasets. However, these evaluations are often confined to homogeneous environments, such as one repository or specific project versions. Consequently, it remains unclear how well these tools generalize beyond their originally targeted scenarios. We present Porting Benchmark, a curated dataset of 1,234 security patch backporting cases spanning cross-version, cross-branch, and cross-repository scenarios, paired with a common evaluation framework. Using this benchmark, we evaluate five tools spanning program analysis, LLM prompting, and LLM agents under aligned settings. Our results show that aligned evaluation changes the apparent performance landscape: PortGPT and TSBPort remain comparatively strong on the Replication Dataset, while FixMorph and Mystique degrade substantially under the common protocol. Performance degrades sharply on structurally complex patches: the best commit-level success rate falls from 85.2% on Type-I patches to 24.0% on Type-IV. We identify four root-cause categories (missing target API awareness, cross-version semantic mismatch, non-local dependency propagation failure, and patch construction or localization failure) and derive concrete directions for next-generation tool design. On a 45-case dynamically validated subset with verified test cases and constructed POCs, we further observe that reference-based benchmark scores do not fully capture real-world remediation: exact match sharply under-credits harder target adaptations, while executable validation reveals residual integration failures in the target that static reference agreement misses. Executable-feedback refinement provides limited but measurable recovery on the hardest executable cases.

LLM-driven agents can autonomously exchange opinions on online platforms and form communities. Such agent-operated social platforms raise a new security concern: attackers may manipulate agents to induce group polarization. Existing methods manipulate agent prompts or construct echo chambers, both of which are difficult to realize in practice. We therefore formulate a new threat, Memory-Mediated Polarization Cascade, which uses agent memory as a persistence channel and public discussion as a propagation channel. This threat contains three stages. During exposure and memory retention, the attacker exposes a small set of target agents to arguments that reinforce their respective stated stances. The targets' memory systems then process and retain these arguments. During retrieval and reproduction, a shared stance-neutral discussion cues the targets to retrieve and reproduce their respective retained arguments. During iterative propagation, untreated agents influenced by the reproduced arguments restate and spread them. We instantiate this threat in GraphWake with three components: (i) stance-support argumentation knowledge graphs construct knowledge-based arguments; (ii) axiom-oriented triple selection distills them for reliable retention and reproduction; and (iii) stance-neutral memory cueing triggers concurrent retrieval and reproduction, initiating propagation. Experiments across multiple discussions and memory systems show that GraphWake substantially increases group polarization. These findings reveal a community-level polarization risk.

LLM-powered GUI agents that autonomously operate smartphones are rapidly transitioning from research prototypes to early real-world deployment. However, because these agents routinely process untrusted environmental content, they are highly vulnerable to environmental injection attacks, which include indirect prompt injections and adversarial instructions. Such attacks can manipulate the behavior of agents without user awareness through diverse channels encountered in everyday mobile use. Despite these risks, existing benchmarks often fail to capture everyday user scenarios, lacking a systematic evaluation of GUI agents under environmental injection attacks on mobile devices. To address this gap, we introduce MobileWorldSafety, a benchmark of 142 risk tasks built on real Android applications. For each task, we define a programmatically verifiable risk indicator over the final system state and evaluate outcomes with a two-stage pipeline: rule-based verification handles unambiguous cases, while an LLM judge adjudicates ambiguous ones. This distinguishes safety failures from capability failures and enables objective and reproducible assessment. Evaluations on six agents, including both general agents and specialized GUI agents, demonstrate that all agents remain highly vulnerable, with attack success rates ranging from 40.4% to 66.9%. These findings indicate that current agents often fail to maintain safety alignment when adversarial content is presented as ordinary mobile context. MobileWorldSafety provides a foundation for quantifying these vulnerabilities and advancing research on robust mobile GUI agents.

We present rl-triton, an open-source library of high-performance GPU kernels for reinforcement learning credit assignment, implemented in Triton. The core contribution is a unified associative scan framework that recasts seven distinct RL estimation algorithms - Generalized Advantage Estimation (GAE), V-Trace, Retrace($λ$), TD($λ$) returns, discounted returns, eligibility traces, and episodic prefix sums - as instances of a single first-order linear recurrence solved in $O(\log T)$ parallel steps. All algorithms share the same associative scan operator, with algorithm-specific fused Triton kernels constructing their recurrence coefficients on-chip. We verify the associative operator algebraically and define the treatment of terminated and truncated episodes explicitly. Benchmarks show a 1.6-5.70$\times$ full-call speedup over a vectorized torch.compile baseline in the massively parallel simulation regime (thousands of environments, short rollouts). The reported range covers all seven algorithms on both GPUs, both with and without per-step truncation handling. For most algorithms, speedups increase at longer sequence lengths, as the baseline requires more scan stages as $\log T$ grows, each adding an intermediate HBM round-trip. The library is available at https://github.com/simonsays1980/rl-triton.

Vision-LSTM (ViL) enables efficient global modeling, but its cost still scales with the number of spatial tokens, so existing segmenters confine ViL to a coarse bottleneck and lose fine anatomical detail. Rasterizing 2D features into a 1D sequence further breaks adjacency across the orthogonal scan axis. We propose MaLViL, a Multi-axis Low-rank Vision-LSTM network that extends ViL across decoder resolutions. Bidirectional low-rank ViL (Bi-LRViL) reasons on a compact orthonormal subspace and preserves detail through an orthogonal residual; scale-aware SaLViL restores cross-axis neighbors before serialization; and a Cross-Directional Mixer (CDM) fuses orthogonal horizontal and vertical traversal paths. Statistics-Guided Skip Modulation (SGSM) further retains boundary cues in encoder skips. On skin-lesion, ultrasound, and multi-organ CT benchmarks, MaLViL achieves competitive or state-of-the-art segmentation accuracy, while reducing ViL operator memory by up to $83\times$ at fine decoder resolutions. Code is available at: https://github.com/xmindflow/malvil.

Integrating open-vocabulary perception into object-level 3D scene graphs is a double-edged sword. While vision-language detectors recover long-tail categories and small, fine-grained objects overlooked by closed-set models, they also tend to fragment large surfaces and merge small objects into larger neighboring objects, compromising instance-level consistency and undermining mapping fidelity. Moreover, existing methods struggle to retrieve previously unmapped targets or determine whether a queried object is absent, hindering robust embodied open-world navigation and exploration. We present OVIP-SG, a unified framework for instance-preserving semantic mapping, functional scene partitioning, and language-guided small, fine-grained object retrieval. OVIP-SG uses a vision-language model (VLM) to enumerate scene-specific categories for robust open-world detection. Symmetric 3D Intersection over Union (IoU) association and area-weighted feature fusion preserve small independent instances, while VLM-inferred object functions partition scenes into compact functional search regions. A four-stage cascaded retrieval pipeline further incorporates voxel voting and determines target absence from exploration coverage. Under a unified evaluation protocol on Replica, OVIP-SG outperforms ConceptGraphs by 6.31 points in class-mean accuracy (mAcc) and 5.15 points in frequency-weighted mIoU (F-mIoU) while achieving a class-agnostic native-instance Panoptic Quality (PQ) of 0.398. It reduces the search area to 21.8% of the indoor floor space and reaches 0.773 balanced accuracy for object-presence classification. Real-world robotic experiments further demonstrate its practical effectiveness.

Large language models (LLMs) can both expand underspecified queries and encode text as dense representations, suggesting a unified model for query expansion and retrieval. Existing systems usually rely on prompted expansions, independently trained modules, or staged optimization, leaving generated expansions only indirectly aligned with the retrieval loss that judges them. We train a single decoder-only LLM end to end, where the same model generates the expansion and encodes both the expanded query and candidate documents. This unified setting creates a moving-target problem: retrieval supervision should improve query-side expansion, but the same update also shifts the document embeddings that serve as retrieval targets. We introduce Document Embedding Preservation Tuning (DEPT), which keeps tuned document embeddings close to cached initial embeddings while allowing retrieval gradients to pass through straight-through decoding into the generator. DEPT converts joint query--document movement into query-side adaptation against approximately stable, whitened document embeddings that support index reuse and online hard-negative mining. Experiments with Qwen3-4B-Instruct-2507 and LLaMA-3.2-3B-Instruct on five datasets in BEIR benchmark show that DEPT improves average retrieval quality over training-free, independently trained, and staged unified baselines, while ablations isolate the effects of preservation, whitening, end-to-end expansion training, and online negatives. Code is available at https://github.com/ILSparkle/DEPT.

Developing robots capable of understanding and manipulating objects requires compact, interpretable, and generalizable representations. This work proposes a reinforcement learning-based framework for robotic grasp refinement, integrating keypoint-based object representations with a Deep Q-Network (DQN). Using 2D overhead images captured in a simulated environment, a geometric-based algorithm generates initial grasp candidates, which are iteratively refined by the proposed framework, transforming failed grasps into successful ones. Experiments conducted on 300 objects from the Dex-Net dataset using a UR5 manipulator demonstrate the framework's effectiveness, achieving a 100% success rate on objects previously deemed ungraspable by geometrical methods. The framework's sim-to-real transferability is further validated through physical experiments on a Delta parallel robot, where a refined grasp successfully manipulates an object that was previously ungraspable. The findings underscore the effectiveness of reinforcement learning in addressing challenges in robotic grasping, offering a scalable and adaptable solution for contact-rich manipulation tasks.

Machine learning models for electroencephalography (EEG) analysis show great promise across a wide range of applications, but their deployment in high-risk domains is hindered by their vulnerability to distribution shifts. Encountering out-of-distribution (OOD) data can lead to catastrophic, overconfident predictive failures. While OOD detection methods can mitigate these risks, they remain heavily under-explored for EEG. Moreover, evaluations in the broader literature typically evaluate OOD detection performance in isolation, ignoring their practical impact on downstream applications. To bridge this gap, we introduce a benchmark for EEG OOD detection, evaluate a broad range of methods, and furthermore evaluate their value in two clinical downstream prediction task. Our results disentangle OOD detection and model uncertainty estimation capabilities, which are frequently conflated in the literature, provide actionable insights about the current state of the art for EEG OOD detection and model uncertainty estimation, and demonstrate how complementary methods for both aspects can be combined to form a robust safety net for the deployment of EEG-based machine learning models in real-world applications.

We present MoNe, a lightweight modular neural memory that attaches to any frozen pretrained Transformer to enable long-context inference without retraining. MoNe reads context in fixed-size segments via test-time learning of fast-weight neural memory networks with layer-localized gradient updates; at inference, the memory generates keys and values from the query tokens alone, with no context tokens re-read. This two-phase design decouples inference cost from context length, achieving $O(N)$ preprocessing and $O(1)$ query cost with peak GPU memory that does not grow with $N$. At 128K tokens, MoNe reduces both compute and peak GPU memory by approximately 80% compared to ICL with only 6.4% parameter overhead. MoNe generalizes to context lengths far beyond the backbone's native window, achieving strong performance on needle-in-a-haystack and word extraction benchmarks from RULER, where ICL degrades sharply.

We consider the dynamic principal-agent problem under asymmetric information, wherein a principal sequentially designs contracts to incentivize an agent with unknown preferences and hidden actions. A fundamental bottleneck in the existing literature is the assumption of deterministic agent utility, which renders the principal's expected utility discontinuous and forces computationally intractable discretizations of the contract space. In this paper, we address this limitation by introducing a stochastic counterpart into the agent's utility model, capturing the inherent physical and behavioral variations in realistic subsystems. We formally prove that this stochastic formulation restores the continuity of the principal's expected utility. Leveraging this continuous geometric structure, we formulate the interaction as a structured multi-armed bandit problem subject to heteroscedastic noise. We propose a \texttt{Heteroscedastic GP-UCB} algorithm that utilizes a Neural Network (Arcsin) kernel, chosen to capture the non-stationary, sigmoidal geometry of the utility landscape. For an $m$-dimensional compact contract space, we establish a high-probability cumulative regret bound of $O\left(\sqrt{T}(\log T)^{m+1}\right)$. Finally, we demonstrate the practical efficacy of our theoretical framework by formulating the Vehicle-to-Grid (V2G) incentive design problem, proving its equivalence to a dynamic principal-agent problem, and showing superior economic performance for grid aggregators.

Whole-slide pathology reasoning requires models to integrate gigapixel-scale visual evidence across complete case-linked slides, yet current question-answering benchmarks primarily measure final answer accuracy--a metric vulnerable to linguistic priors and benchmark regularities, and insufficient to establish that predictions are grounded in the supplied tissue. We introduce PathoArgus-Bench, a benchmark and evaluation protocol that explicitly tests the full evidence chain: availability, accessibility, use, and responsiveness. PathoArgus-Bench comprises 22,078 four-choice questions from 4,913 patients across 15 TCGA projects, covering six pathology capabilities across three levels of evidence demand, and operates under a fixed reader budget that retains only a small fraction of the gigapixel context. To further isolate evidence-grounded reasoning, we contribute ESG (Evidence State Quartets), a controlled set of 483 quartets where the question text is fixed while the target WSI set is moved, replaced, or removed, requiring consistent predictions across all states. Evaluating 20 general-purpose, medical, and pathology-specific systems reveals a stark gap: while GPT-5.6 achieves 57.09% overall accuracy and 57.04% on ESG, it correctly completes only 19 of 483 quartets (3.93% QExact), exposing that row-level accuracy does not translate into reliable evidence grounding. We also introduce PathoArgus, a fixed-budget reader that allocates context via question relevance and spatial coverage, attaining 50.39% overall accuracy yet only 1.86% QExact--demonstrating that improved context access alone does not ensure consistent evidence-based prediction. Our benchmark and diagnostics establish that acquiring useful whole-slide context is necessary but far from sufficient, and call for a shift from answer-centric to evidence-grounded evaluation in computational pathology.

Conversational AI is moving beyond isolated text prompts toward sustained, multimodal interaction. In real conversations, users clarify goals, revise requests, interrupt responses, switch topics, and introduce new evidence while expecting systems to preserve context across turns. This makes multi-turn dialogue a distinct challenge requiring systems to maintain and update memory, ground responses across modalities, tools, and external knowledge, and adapt across languages and cultures. This study reviews multi-turn conversational AI across text-only dialogue, AudioLLMs and speech-native systems, multimodal and omni-modal systems, and tool-augmented agents. We organize the literature around datasets and benchmarks, modeling paradigms, training strategies, evaluation setups, and cross-cutting challenges. Our analysis shows that support for multiple modalities has advanced faster than the ability to sustain coherent interaction across a session. Despite stronger capabilities to perceive, speak, and act across modalities, current systems still struggle with persistent memory, cross-turn grounding, full-duplex interaction, robust evaluation, and cultural alignment. We conclude with a research agenda for systems that can remember, revise, ground, speak, listen, act, and adapt across turns, modalities, and cultures. (https://github.com/faiza-sfa/multiturn-conversational-ai-survey)

This paper studies tactile-only 6-DoF pose refinement and belief maintenance for grasped objects in static and short quasi-static in-hand configurations where vision is unavailable or heavily occluded. The key difficulty is tactile partial observability: whole-hand taxel contacts are sparse, intermittent, and ambiguous under limited excitation and object symmetries. We propose a physics-informed particle filter on $\mathrm{SE}(3)$ that updates pose beliefs from dense whole-hand tactile measurements. The likelihood combines active-contact signed-distance consistency, force-normal alignment, friction-cone feasibility, zero-force negative evidence, and optional feasibility guards. A sliding-window log-likelihood fuses recent tactile frames to reduce single-frame ambiguity, while a potential-field-guided proposal steers particles away from hand--object penetration. Symmetry-aware resampling preserves multiple plausible modes. Experiments on an Allegro Hand V5 with five objects show lower normalized ADD-S than tactile-only geometric, particle-filter, and learning baselines, and ablations confirm the benefits of temporal fusion, potential guidance, and mode preservation.

Visual cues are increasingly adopted to guide robot learning, but whether Vision-Language-Action (VLA) models can reliably follow authorized cues while disregarding unauthorized ones remains unclear. Existing work covers only a narrow range of cue forms and focuses on final task success, providing only a coarse assessment of cue-following capability. Treating all visual cues as authorized also leaves safety risks of unauthorized following unexplored. To address these gaps, we introduce LIBERO-VIFO, a benchmark to evaluate both the capability and safety of visual cue following in VLA models. LIBERO-VIFO defines eight visual cue families spanning diverse forms. A total of four protocols in two parts are defined: Part I tests cue understanding and authorized following, while Part II evaluates unauthorized visual cue following under language-cue conflict and empty language conditions. Evaluating seven VLA models reveals that although visual cue understanding does not reliably translate into execution, current VLAs are able to execute cue-indicated tasks without language instruction, exposing an emerging risk of unauthorized visual cue following. Extended experiments on scene-instantiated cues, safety-critical settings, and real-robot deployment corroborate these findings. LIBERO-VIFO brings both the capability and safety of visual cue following into systematic evaluation, establishing visual-centric safety as a new perspective for the VLA community.

Large language models are increasingly deployed through agent harnesses that manage tools, extensions, persistent state, permissions, and external actions. Existing safety benchmarks mainly target individual attack mechanisms or a limited subset of operational settings, making it difficult to compare how safety failures emerge across different harness responsibilities. We present HarnessRisk, a lifecycle oriented benchmark that organizes agent harness safety into six operational phases including Harness Configuration, Capability Extension, Runtime Operation, State Persistence, Action Control, and Incident Recovery. HarnessRisk contains 128 sandboxed cases, each pairing a benign user objective with an adversarial instruction embedded in an untrusted workflow artifact. We evaluate each trajectory using Utility, Attack Success Rate, Persistence, and Detection. Across three harnesses, six language models, and 14 model and harness configurations, attack success ranges from 12.6% to 80.9%, while Utility remains between 75.0% and 97.6%. Harness Configuration is the most vulnerable phase across all three harnesses, showing that attacks can succeed by altering security sensitive parameters within otherwise authorized workflows. We also find that explicit risk recognition does not reliably lead to safe action, as some configurations detect risks in more than 90% of runs while retaining substantial attack success. These results highlight the need to evaluate agent safety across multiple harness responsibilities and at the level of the deployed model and harness configuration.

In this study, we investigate developmental mechanisms that enable small, resource-constrained systems such as cm-sized millirobots to autonomously explore, learn, and adapt their capabilities throughout their lifespan. Reinforcement learning algorithms guide the agent's skill acquisition and adaptation through the interplay of our proposed tinyDSM, which integrates intrinsic motivation and fitness-based assessment. We strive for minimal, hard-wired skills while encouraging the open-ended development of new skills. A key emphasis in our approach is to encode minimal a-priori general knowledge, which serves as a foundational starting point for the system as it further learns system-specific dependencies from the initial knowledge provided. Thus, by design, our approach attempts to cover very generic application domains. The methodology is based on (a) developmental mechanism with intrinsic motivation, and (b) a cognitive architecture (knowledge, reasoning, learning), while (c) utilizing minimal resources. It uses a hierarchical knowledge graph and kinematic reasoners to model and evaluate simple and advanced motion related skills. In our experiments, we use a resource-constrained millirobot with a volume of 36 cm^3 with a Raspberry Pi Pico 32-bit microcontroller (RP2040) that integrates all described features and capabilities except the camera system in 9 kB. Starting with learning the most elementary motor skills the millirobot autonomously progresses from simple linear and angular movements to complex geometric patterns within 15 minutes. To complement the physical experiments, we perform a simulation-based analysis that enables systematic comparisons across learning algorithms and intrinsic motivation parameters.

The communication demands of distributed model prediction control (DMPC) can overwhelm even advanced wireless communication technologies as agents must exchange a significant amount of information at least once per time step. To semantically reduce communication demands, this work employs encoder-decoder networks built around long-short term memory (LSTM) cells in a distributed optimization algorithm. Agents publish a reduced representation of a message and receivers reconstruct the original message upon reception. In tests with reduced communication using formations of mobile robots, trained networks retain satisfactory performance and work reliably under conditions overwhelming full communication. As the results show, the usage of LSTMs either allows unprecedented reconstruction accuracy or the usage of different prediction-horizon lengths without the necessity to retrain.

Agent Skills package reusable natural language procedures with executable resources, enabling software agents to acquire task specific capabilities without model adaptation. Automatically generating such Skills can improve task performance, yet evaluating a candidate solely from its artifact or final task outcome leaves unresolved which actions the equipped agent will perform and which side effects those actions will produce. We present TRUSS, an evidence guided framework for generating functionally effective and safety reliable Agent Skills. TRUSS first inspects functional claims against source and domain evidence while evaluating the complete artifact under nine predefined safety properties. Candidates admitted by this static gate are loaded by a shadow agent inside a Controllable Execution Environment, where brokered tools expose requested actions to policy enforcement and record their results as provenance preserving execution traces. Functional failures and property violations are linked back to the responsible Skill content and used to guide iterative refinement. We evaluate TRUSS on 168 SkillInject artifacts, 155 SkillSafetyBench cases, and all 187 tasks in SkillGenBench. TRUSS achieves 100.00\% precision and recall in vulnerability detection. Repair reduces attack success from 38.71\% to 19.35\% with GPT 5.5 and from 46.45\% to 29.68\% with GPT 5.4, with zero attack regression. For Skill generation, TRUSS raises task effectiveness from 17.11\% without Skills to 52.94\%, while increasing the benchmark Security rate from 50.80\% to 100.00\%. These results show that execution evidence can expose behavioral failures missed by artifact inspection and can guide Skill generation toward jointly verified functional and safety outcomes.

Expert-written natural language skills can improve tool-using agents, yet agent-authored skills perform 8-11 points worse than using no skill. This gap suggests that following procedural guidance and improving it from execution evidence are distinct capabilities. Inference time loops can repair skills but do not improve the model that writes the next one. We study how to organize execution experience from intermediate skills into training states for an optimizer. We introduce WER (Write, Execute, and Refine), a multi-phase framework that trains a Skill Optimizer outside a frozen executor. The optimizer proposes skills, a frozen agent executes each repeatedly, and a programmatic verifier scores the outcomes. The scores provide relative credit and select mixed-outcome records. Matched successful and failed trajectories from these records form the next phase's refinement states, so the optimizer learns from the consequences of its earlier outputs. On BFCL v4 multi-turn and tau2-bench, WER improves average Pass@1 over the no-skill baseline by 7.80 and 3.85 points, respectively. Under an identical refinement workflow, it outperforms the same backbone without optimizer training by 9.35 and 10.29 points. The trained 4B optimizer reaches 76.63 percent on BFCL v4, outperforming all evaluated off-the-shelf general-purpose models used as skill optimizers on average.

We propose HODAgent, a System-2 embodied agent for humanoid robots in service settings, addressing situated intent, responsive execution, task revision, and outcome verification. Its semi-duplex architecture integrates an Env-Interactor, Planner, Executor, and hierarchical Memory to maintain coherent interaction, planning, and task state during service episodes. This allows handling new requests during motion, retaining progress, revising actions, and grounding closure in execution outcomes. A shared interface connects simulation and physical robots (Unitree G1), isolating platform-specific control. In an interactive simulation with 164 cases, HODAgent achieves 84.8% and 91.5% Joint Success under two VLM backbones, outperforming baselines by 9.8 and 18.9 points. On physical robots, pass rates are 92% (atomic), 72% (composite), and 63.3% (complete tasks). On multiple embodied benchmarks, it improves over baselines by 0.7-9.0 points. Results show a unified System-2 agent enables adaptive humanoid service across simulation and reality.

How cautious should an agent be while it is still learning its environment? We propose RATTL (Risk-Adversarial Total-Reward Learning), which ties caution to epistemic uncertainty: the agent holds a Bayesian posterior over unknown dynamics and plans against a Wasserstein ambiguity set whose radius is a monotone function of that posterior. The radius contracts with evidence, so behaviour interpolates continuously between worst-case robustness and risk-neutral total-reward maximization. The design follows the duality underlying the Entropic Value-at-Risk, which converts the choice of a risk level into the choice of an ambiguity radius. We show the resulting planning problem is well posed under transience and compactness conditions, and prove a Safety Sandwich: the RATTL value lies between the uninformed robust value and the full- knowledge optimum, with a gap that vanishes as the posterior concentrates. In a canonical binary-hazard instance, the induced criterion reduces to Conditional Value-at-Risk at a level set by the posterior entropy. A worked example shows the agent deferring the efficient action until a sharp identification threshold. RATTL targets runtime safety for agents, including LLM-based systems, acting under uncertainty.

Motivation: Low-dimensional embeddings are widely used to explore cell-state heterogeneity in single-cell and other high-dimensional biological data. Although many methods preserve local neighborhoods, they may distort the apparent sampling density of processed observations, altering the visual contrast between dense and sparse regions and complicating the interpretation of rare, transitional, or continuous cell-state populations. Results: We present DMT-Dens, a parametric manifold-visualization method built on a latent-token Transformer encoder. The model integrates rank-based manifold alignment with hard-pair aggregation. To preserve density, it optimizes a loss based on the Pearson correlation between k-nearest-neighbor log-radius estimates in the processed input and two-dimensional embedding spaces. Benchmark evaluations demonstrate strong density preservation, particularly on biological datasets, while retaining competitive label separability. Availability: Source code, data-processing scripts, and resolved experiment configurations are available at https://github.com/Ruizhe-wang/DMT-Dens.

Pretrained molecular language models are increasingly used as molecular encoders for learning structure-property relationships. However, their practical suitability for molecular discovery within and beyond their pretraining domain remains unclear. Herein, we systematically benchmark four molecular language models across six virtual molecular libraries spanning drug discovery, organic materials, and catalysis. Native molecular language model embeddings show substantial variation in discovery performance across libraries, whereas molecular fingerprints provide a consistently strong and robust baseline. Consistent with a potential domain-representation mismatch, we show that explicit domain adaptation substantially improves representation performance. Fine-tuning molecular language model encoders on structures from the target virtual library consistently improves sample efficiency, with several adapted encoders emerging as the top-performing representations across the benchmark tasks. These results show that molecular representation quality depends strongly on the target domain and that explicit adaptation can improve the practical utility of molecular foundation models. More broadly, our findings establish domain-adapted molecular representations as a promising strategy for sample-efficient adaptive decision making in virtual screening and self-driving laboratories.

The quality and diversity of instruction-based video editing datasets are steadily improving, yet existing datasets mainly focus on single editing operations and fall short in supporting compositional instruction-guided video editing. In particular, multiple editing intents must be jointly understood and faithfully executed within the same video. To address this issue, we introduce CoinVE-200K, a large-scale, high-quality dataset for Compositional Instruction-Guided Video Editing. CoinVE-200K contains 1080p video-editing pairs of up to 201 frames, covering diverse compositional scenarios where each sample involves 2 to 5 atomic editing operations. The instructions target humans, objects, and backgrounds, and cover edit types such as addition, removal, modification, and stylization. All samples are built through a carefully designed generation and filtering pipeline to ensure instruction faithfulness, visual quality, temporal consistency, and compositional diversity. We also introduce CoinVE-Bench, a benchmark for compositional-instruction video editing across diverse subjects, operation types, and instruction complexities. Furthermore, we present CoinVE-Edit, a 22B compositional video editing model built upon Wan2.1-T2V-14B and Qwen3-VL-8B-Instruct. CoinVE-Edit disentangles region-aware attention for different editing instructions, enabling precise multi-region editing while preserving irrelevant content and temporal coherence. Experiments on CoinVE-Bench show that CoinVE-Edit achieves strong performance in instruction following, compositional editing accuracy, visual quality, and temporal consistency.

The health of marine ecosystems is a critical indicator of global environmental change, yet the physical constraints of underwater observation and the intrinsic challenges of processing marine imagery severely limit the scalability of systematic monitoring. While recent visual foundation models such as the Segment Anything Model (SAM) series show great promise, they still struggle with the fine-grained recognition required in these complex scenarios and still require expert supervision. Our work addresses this gap by bridging state-of-the-art foundation models with existing sparse supervision. Because historical benthic surveys are typically annotated with only a few sparse expert points per image, we utilize these legacy point-labels as visual prompts for SAM2. Our primary contribution is a novel mechanism to automatically identify which of these points are suitable, and which are actively harmful, when used for propagation. By filtering out unreliable points, we extract high-quality pseudo-ground-truth masks capable of training more accurate, fine-grained semantic segmentation models. We demonstrate the effectiveness of our approach on public benthic data and introduce a new, challenging benchmark featuring real-world sparse expert annotations, paving the way for scalable ecological analysis.

In this paper, we tackle the problem of performing consistent, unified modifications to a multi-shot video sequence. This task is particularly challenging because multi-shot videos consist of discontinuous temporal segments that vary significantly in viewpoint, camera scale, and subject pose, leading to severe identity drift and cumulative error propagation. Achieving coherent edits requires establishing reliable cross-shot semantic awareness to maintain stable subject appearance and visual continuity across these disjointed boundaries. To address this, we propose MSEditor, the first framework designed specifically for consistent multi-shot video editing. To overcome the scarcity of high-quality multi-shot training data, we repurpose existing multi-view video datasets to provide robust cross-shot supervision. Architecturally, we introduce a Supervisory Adapter that injects this cross-shot information into the diffusion backbone, enabling the model to learn identity-consistent representations. Furthermore, to effectively mitigate cumulative errors and ensure long-range temporal coherence, we design a Cross-Shot Packing strategy that dynamically aggregates information from semantically related shots within the self-attention window. Extensive experiments demonstrate that MSEditor significantly outperforms existing methods on our curated multi-shot video editing benchmark in terms of identity preservation, temporal stability, and overall visual quality.

Cameras are ubiquitous sensors in robotics due to their compact form factor and the perceptual richness captured through visual information. Monocular SLAM enables robots to understand the environment with a minimum setup, however, it inherently suffers from scale ambiguity. A common solution is to provide multi-modal sensor configurations, such as visual-inertial systems, where scale is observable unless the robot navigates under a constant-velocity motion, a common scenario in mobile robotics. With the advent of deep-learning, geometric foundation models have been used to address this problem, but the depths maps are often noisy and scale-inconsistent across frames. In this paper, we propose Scalix, a real-time monocular SLAM framework that achieves metric-scale state estimation by integrating learned depth cues into a probabilistic factor-graph formulation. By augmenting existing monocular depth models with both per-pixel depth uncertainty and per-frame scale uncertainty, Scalix treats scale predictions as independent measurements within its optimization, leading to improved scale consistency through multi-view data associations. Experiments in large-scale outdoor and indoor environments demonstrate state-of-the-art performance on both metric and up-to-scale benchmarks while maintaining real-time operation and generalization.

Academic papers are a primary carrier of scientific knowledge, yet most of this knowledge remains locked in PDFs that are optimized for human reading rather than machine use. For Multimodal Large Language Models (MLLMs), the core challenge is not only perception, but representation: scientific pages interleave text with Structured Academic Elements (SAEs) such as tables, formulas, charts, and pseudocode, whose structure, data, and logic are poorly preserved by common surrogates like Markdown. We therefore propose Compilable Academic Document Parsing (CADP), a paradigm that reconstructs a full page as contextual \LaTeX{} plus executable Python, so that structure-preserving elements and executable chart representations can be reconstructed, recompiled, and directly verified against the source page. To support this setting, we introduce CADP-Bench, an expert-verified benchmark of full academic pages containing tightly coupled text and multiple SAE types, evaluated through a re-injection compilation protocol. We further study current capabilities using SOTA MLLMs and an exploratory multi-agent baseline that incorporates common agentic techniques. Results show that even frontier models still struggle to produce high-fidelity executable reconstructions, highlighting substantial room for improvement in structure-aware scientific document parsing. CADP-Bench is released for future research.

Joint-Embedding Predictive Architectures (JEPAs) learn world models by predicting future embeddings, but the objective admits a trivial solution of a constant encoder, so every practical system adds an anti-collapse mechanism (LeCun, 2022; Assran et al., 2023; Bardes et al., 2022; 2024). LeWorldModel (LeWM) prevents collapse with SIGReg, a regularizer that forces the latent distribution to match an isotropic Gaussian: the representation is stabilized by prescribing what it must look like, independently of the environment it models. We argue that the anti-collapse pressure can instead come from the transition data itself. Action-Contrastive Masked Transition Modeling (AC-MTM) keeps LeWM's forward latent-prediction objective and adds a training-only inverse-dynamics head trained with Action-NCE: each latent transition must identify the action that produced it among the other actions in the batch, a discrimination task that a collapsed encoder provably fails. The inverse branch is discarded after training, leaving test-time encoding, forward prediction, planning, and compute identical to LeWM. On four standard pixel-control tasks under a matched planning protocol, AC-MTM trains stably from scratch and matches SIGReg on average. On the harder multi-object OGBench Visual Scene task, results are consistent with the prescribed geometry becoming a bottleneck: AC-MTM reaches 80.0$\pm$2.0% success versus 58.0$\pm$2.0% for SIGReg, improving by 20-24 points in each training seed. A single 50-episode random-policy run gives a 52% baseline estimate. Contrastive inverse dynamics thus provides a distribution-free anti-collapse signal that requires no target network, stop-gradient, pretrained encoder, or reconstruction objective, and we characterize the action-space and observability assumptions under which it holds. We make our code available at https://github.com/jackboyla/action-contrastive-jepa

Legal consultation questions exhibit multi-level complexity. A single retrieval strategy often leads to over-reasoning for simple questions and poor interpretability for complex ones, making it difficult to meet the requirements for both answer quality and efficiency in high-risk scenarios. To address this issue, this paper proposes CoAL-RAG, a complexity-aware legal retrieval-augmented generation method, which constructs a multi-dimensional evaluation mechanism based on ``question essence'' and ``retrieval consistency'' to enable adaptive routing of retrieval strategies. First, the reasoning demand is quantified according to the logical structure of the question. Then, the discrepancy between semantic retrieval and keyword retrieval is utilized to indirectly reflect problem complexity, thereby selecting the most appropriate retrieval strategy and dynamically filtering contextual information. Experimental results demonstrate that the proposed method significantly outperforms baseline models not only on Chinese legal benchmarks (SocialLawQA, LawBench) but also demonstrates strong cross-jurisdictional generalization on English datasets (LexGLUE, CaseHold). Specifically, on Chinese datasets, the BLEU score improves by 42.5\% and ROUGE-L reaches 3.6 times that of knowledge graph-based methods. On English benchmarks, CoAL-RAG maintains highly competitive accuracy, achieving an optimal balance between generation quality, deep logical reasoning, and system efficiency across different legal systems.

Simultaneously reconstructing and understanding 3D environments is essential for embodied agents. Toward this goal, feed-forward semantic 3D Gaussian Splatting (3DGS) efficiently constructs semantic scene representations from sparse multi-view observations. However, existing methods lack explicit instance discrimination and mainly support category- or phrase-based semantic queries. To this end, we propose GroupForward, an instance-grouped feed-forward Gaussian splatting model that reconstructs geometry, appearance, instance structure, and semantics from sparse, unposed, and uncalibrated multi-view images. Unlike existing methods that attach high-dimensional semantic features to each Gaussian, GroupForward learns compact instance embeddings that group Gaussians into cross-view consistent 3D instances, reformulating feed-forward semantic 3DGS from per-Gaussian semantic feature rendering to instance-level semantic aggregation and propagation. Building on these instance groups, we further propose a Referential Scene Reasoning Framework (RSRF) for complex 3D referring segmentation. RSRF constructs an instance-grouped 3D scene graph and retrieves candidate instances for a given referring expression. A vision-language model then reasons over structured instance evidence and multi-view observations to identify the referred instance among the candidates. RSRF thereby extends language interaction from simple semantic querying to complex referential scene reasoning. Experiments on semantic reconstruction and referential reasoning demonstrate the effectiveness of our instance-grouped reconstruction and reasoning framework.

Large language models increasingly serve as persistent conversational assistants, requiring memory that preserves relevant experience and maintains continuity across interactions. Existing methods improve access to conversational history through long-context processing, selective retrieval, and structured memory organization. However, most systems treat memory access as retrieving relevant past information without first determining which prior interaction state the current turn resumes. This limitation becomes particularly important when conversations interleave multiple tasks, people, and plans that may be interrupted and later revisited. We introduce ArborMem, an online memory framework that represents a long-running conversation as a navigable forest of interaction states. Each branch preserves a locally coherent trajectory, while the forest maintains multiple trajectories that may later be resumed. For each new input, ArborMem localizes the relevant state, restores its branch-local context, and augments it with reusable evidence retrieved across branches, preserving interaction continuity without conflating semantically related but structurally distinct trajectories. Existing long-term memory benchmarks cover diverse memory and reasoning capabilities but do not explicitly isolate branch-structured challenges. We therefore introduce BranchMemEval, a controlled diagnostic benchmark for interleaved and resumable interaction trajectories. Experiments on LongMemEval, LoCoMo, BEAM 100K, and BranchMemEval show that ArborMem outperforms the strongest baselines by 3.36 to 10.31 percentage points on the three established benchmarks and by 5.0 points on BranchMemEval. Its advantage grows under constrained read budgets, while complete memory queries remain below half a second.

Continual pre-training of large language models must acquire new information without erasing old knowledge. Existing replay methods often choose a global old/new mixture and sample uniformly, ignoring that examples differ in how quickly they are forgotten. We formulate continual pre-training as adaptive review scheduling: the training loop should decide not only how much history to replay, but which examples should return at each step. We introduce Spaced Repetition Training (SRT), a continual learning framework inspired by cognitive science, which schedules sample-rehearsal using the SuperMemo-2 (SM-2) algorithm. SRT maintains per-example review state, maps per-example perplexity to a recall-quality signal, and schedules historical examples for retention and new examples for consolidation while leaving the model, objective, and optimizer unchanged. On temporally separated Wikipedia and code corpora, SRT improves the stability-plasticity trade-off, recovering 5 to 37 percentage points of old-knowledge accuracy lost by naive continual pre-training across model scales while preserving or improving new-knowledge acquisition. At larger scale, SRT preserves broad benchmark performance that naive continual pre-training and uniform replay substantially degrade. Experiments with vision and tabular data further suggest that the scheduling principle extends beyond language when paired with an appropriate recall signal.

Modern agents operate inside agent harnesses that manage tools, context, and control flow, making the harness a critical part of the agent system. Our original Agent Lightning introduced a disaggregated architecture that connects arbitrary agents to RL training through an LLM endpoint proxy, an approach later adopted by frameworks such as verl Uni-Agent, AReaL 2.0, slime, and Polar. We refer to this paradigm as harnessed agentic RL, where the deploy-time harness directly participates in model post-training. Harnessed agentic RL differs fundamentally from traditional agentic RL: the harness, rather than the training engine, owns the environment interaction loop, while the trainer observes only sequences of LLM request-response pairs. This introduces challenges in retokenization, sample merging, advantage calculation, loss normalization, and backend scheduling, which can substantially affect training stability and effectiveness. We present Agent Lightning v1.0, a lightweight framework for harnessed agentic RL implemented in approximately 3,500 lines of code. It supports arbitrary agent harnesses and serves as a practical testbed for studying these challenges. We evaluate it on instruction-following, search, and coding agents, and provide a complete reproducible pipeline for coding-agent RL. Using only 6K training examples and modest compute, RL improves Qwen3.5-9B on SWE-bench Verified from 41.8% to 56.4%, a 14.6-point absolute gain. We release the complete workflow and training scripts to facilitate reproducible research on harnessed agentic RL.

This preliminary paper outlines a planned evaluation benchmark for Explainable Reinforcement Learning (XRL) methods. Current evaluations rely on functionally-grounded metrics like faithfulness and compactness, and on human-grounded proxies like subjective ratings or prediction accuracy. We suggest evaluating XRL methods by how effectively their generated explanations help to diagnose and fix malfunctioning reinforcement learning (RL) agents. We propose EvalXRL, a benchmark in which a Large Language Model (LLM) coding agent uses different XRL methods to diagnose a held-out malfunction in an RL agent, and then repair it. Our proposed benchmark iterates across (environment $\times$ malfunction $\times$ XRL method) tuples and uses the reward signal of the RL agents to form a final score for each XRL method. The coding agent may use the method interactively: invoke the XRL method, process its output, form new hypotheses on what is broken, and invoke the method again with parameters adjusted for testing these hypotheses. This closed-loop structure may be described as a simplified version of the scientific method. Some XRL methods provide self-evaluations that follow this pattern; we propose the first head-to-head comparison of multiple XRL methods in closed-loop usage.

Gender bias or other social biases in large language models (LLMs) are frequently evaluated with question answering or survey benchmarks where the LLM needs to give a response in a predefined answer format. It is well known in survey science that the answer format has a substantial impact on answers, just as LLMs are sensitive to the prompt wording. However, to our knowledge it has not been studied yet how changes in answer format impact the measurement of gender bias in LLMs and their alignment with human response distributions. We evaluate three instruction-tuned models on the BBQ benchmark and OpinionQA survey data across closed-ended, Likert-scaled and open-ended formats, comparing bias measurement and distributional alignment under otherwise identical conditions. We find that answer format does substantially alter measured outcomes, including reversals in order rankings. These differences arise because each format elicits distinct response behaviours, such as forced-choice selection, scale-based distributions and refusal in free-text generation. Our findings highlight the importance of treating answer format as a substantive component of LLM evaluation and motivate multi-format designs for more robust model assessment.

Background: Large Language Models (LLMs) are increasingly being applied to Software Engineering (SE) tasks, achieving high accuracy across problems such as clone detection, vulnerability prediction, and code summarization. However, their high computational demands and energy consumption raise sustainability concerns and hinder their use on consumer hardware and resource-constrained platforms. A common way to report the computational cost of an LLM in the literature and industry is to use the number of Floating Point Operations (FLOPs) required to perform a pass over the network. Aims: This paper investigates the implications of energy-aware knowledge distillation for SE, aiming to improve model efficiency while maintaining performance and to determine whether FLOPs is a reliable energy-aware metric. Method: We conduct a controlled experiment using Morph, a Many-Objective Optimization-based distillation methodology, to empirically examine whether FLOPs accurately reflect energy consumption in Clone Detection and Vulnerability Prediction tasks. We extend this methodology to include energy-surrogate models that directly estimate CPU and GPU energy consumption during optimization, and we apply Morph to generative tasks using CodeT5+ for code summarization. Results: Our results show that FLOPs is not always a reliable indicator of energy consumption, and better results can be achieved by using energy-surrogate models. Distilled student models can reduce inference energy consumption by up to 90\% and memory usage by 86\%, with only modest accuracy trade-offs. Conclusions: Energy-aware knowledge distillation when guided by direct energy surrogates rather than FLOPs can improve the energy consumption, sustainability, and deployability of LLMs for SE applications, enabling efficient models on consumer hardware.

Although Large Vision-Language Models (VLMs) have significantly advanced embodied navigation, their direct deployment remains challenging, as existing methods often force VLMs into unnatural action spaces that misalign with their 2D pre-training priors, compounded by rigid reasoning schedules and inefficient memory management. To overcome these limitations, we propose TAMP-Nav, a unified framework for efficient embodied navigation. First, we introduce a Pixel-to-3D Action Formulation (Point) that reformulates navigation into 2D visual prompting. Specifically, the VLM merely selects 2D pixels, which are then projected into 3D coordinates for a low-level SLAM controller. This design naturally aligns embodied execution with the VLM's inherent 2D visual capabilities. Second, we propose an integrated Selective Reasoning and Anchor-Trajectory Memory mechanism (Think and Memorize), which dynamically triggers Chain-of-Thought and retains high-fidelity memory only at critical nodes, compressing redundant trajectories into lightweight Space-Time Indicators, thereby preserving critical historical information and enhancing spatio-temporal perception. Finally, we design an efficient Two-Level Alignment Paradigm (Align) via Group Relative Policy Optimization (GRPO). By superimposing global outcome rewards with fine-grained process rewards, this dense supervision tightly aligns the agent's cognitive planning with physical environmental feedback, endowing the model with adaptive reasoning capabilities. Experiments demonstrate that TAMP-Nav achieves state-of-the-art performance (e.g., 66.2% SR on R2R-CE) with high runtime and sample efficiency (requiring only 90k training trajectories).

Spin defects in silicon carbide (SiC) are promising candidates for integrated quantum photonics, offering long-lived spin states and near-infrared emission suitable for low-loss photonic integration and fibre-based quantum communication. However, light extraction from these defects remains challenging due the relatively high refractive index of SiC. Metalenses offer a compact approach to enhance light collection by engineering the wavefront directly at the material interface. Here, we design and fabricate monolithic metalenses from SiC bulk material that simultaneously operate at 860 and 1240 nm, matching with emission from the nitrogen vacancy and silicon vacancy colour centers. By independently engineering the phase response at both wavelengths, the metalens enables collection and polarization manipulation of the emitted light. We further employ the metalenses to demonstrate optically detected magnetic resonance of both defects simultaneously. These multifunctional metalenses provide a compact optical interface for scalable integrated SiC photonic devices.

Recent efforts toward fully automated AI scientists have demonstrated that language-model agents can generate hypotheses, execute experiments, and draft scientific manuscripts. However, during the early stages of research, when research problems are formulated, these AI scientists often rely heavily on proprietary frontier models. Their proposals are shaped by opaque parametric knowledge and by literature searches conditioned on the proposals themselves. Such knowledge is effectively a black box, and this dependence makes the evidential basis and validity of generated research problems difficult to audit and leaves the process vulnerable to model-specific hallucinations and biases. Furthermore, if proprietary research materials are transmitted to external APIs, the use of these models creates confidentiality, privacy, and data-governance concerns. We introduce the Structural Gap Hypothesis Agent (SGHA), a fully automated, corpus-first research-problem discovery system that runs entirely on a local LLM. SGHA structures a scientific literature corpus into evidence-linked paper objects and a typed evidence graph, detects unresolved structural patterns across papers, screens candidate gaps before formulation, and produces traceable research-problem families. In particular, it is able to output assumptions, objectives, success criteria, and remaining ambiguities. All LLM-based components of SGHA are executed using a locally served open-weight 9B language model, without requiring proprietary frontier-model APIs. We compare SGHA with the AI Scientist-v2 idea formulation module in five machine-learning domains. Our results suggest that explicit corpus structure and evidence-constrained reasoning can support promising, inspectable research-problem formulation without relying on frontier models during generation or verification.

User-facing tool agents must coordinate dialogue and tool use as user goals unfold over multiple turns. Yet interactive reinforcement learning typically reduces each rollout to a terminal reward, assigning the same credit to effective elicitation, errors, and later repair. The next user turn is more than context: it also provides noisy, temporally local evidence about the preceding user-to-user segment. We introduce \textbf{F}eedback-\textbf{A}ware \textbf{C}redit \textbf{A}ssignment (\textsc{FACA}), which aligns each reaction with that segment, derives a locally normalized reaction advantage, and adds it to verified terminal outcome advantage without an extra critic or rollout. Against an outcome-only Interactive GRPO control matched in simulator, visible dialogue, initialization, rollout, and optimization, \textsc{FACA} improves the nine-domain $τ$-family average across three independently trained runs by 5.91 and 10.22 percentage points at 8B and 14B, respectively. Gains concentrate in Telecom; at 8B, randomizing reaction polarity removes the Telecom gain. The same ordering holds zero-shot on Pare-Bench and Co-Gym. These results demonstrate that next-turn user reactions provide actionable local credit for improving multi-turn user-interacting agents.

Vision-language-action policies generalize broadly but provide no execution-time guarantees; classical model-based planners respect kinematic and geometric constraints but generalize poorly. We study whether an action-conditioned Joint-Embedding Predictive Architecture (JEPA) world model can predict, before execution, both task progress and physical risk for candidate action chunks, and whether coupling these predictions to an embodiment-specific model-based safety shield yields a deployable pipeline for heterogeneous robots. We propose a receding-horizon decision pipeline: (1) a proposer produces K candidate action chunks; (2) an action-conditioned JEPA rolls each candidate forward in a frozen-encoder latent space conditioned on an embodiment embedding; (3) calibrated risk and progress heads score each rollout and report uncertainty; (4) a deterministic per-embodiment safety shield filters inadmissible candidates; (5) a fallback ladder handles empty-admissible-set cases. The learned ranking only reorders admissible candidates; enforcement guarantees come from the deterministic shield and fallback ladder. We evaluate with a pre-registered protocol in simulation (LIBERO-Long). In 600-episode configurations the full framework improved success over a shield-only baseline and reduced collision false negatives at matched recall. Deployment-efficiency measurements on target on-robot and edge accelerators are included. Real-robot experiments and an offline reranking significance test remain future work; see the paper for disclosures.

Frozen vision-language-action (VLA) policies are increasingly improved at test time by sampling additional policy behaviors or introducing external demonstrations. Yet there is little guidance for deciding which intervention a deployed policy actually needs. Additional sampling is useful only when better behavior already exists within the policy's stochastic rollouts and can be identified, whereas retrieval is most useful when the relevant action prior is not reliably represented by the policy. We study this decision through two measurable factors, recoverable headroom and retrieval complementarity, which characterize how much useful behavior is already available to recover and whether an external action prior fills a measurable gap. We evaluate an episode-level retry selector under retryable or parallel execution, together with retrieval across multiple frozen VLA policies and environments. The selector consistently recovers substantial latent capability across all tested VLA backbones on LIBERO, with gains of up to 21.0 success-rate points that closely track recoverable headroom. It also transfers to a different robot and simulator and remains effective under degraded observations, while experiments with autoregressive OpenVLA illustrate the distinction between available headroom and the ability to rank candidate rollouts. Retrieval behaves differently, improving the policy with the largest measured action-prior gap and providing further gains when combined with selection. Together, these results provide an empirical basis for characterizing test-time augmentation opportunities by separating capability that can be recovered from the frozen policy from behavioral priors that may need to be introduced externally.

Vision foundation models have recently advanced multi-modal salient object detection (MSOD) through parameter-efficient tuning and prompt learning. However, existing Segment Anything Model (SAM)-adapted MSOD methods often rely on dual-stream encoders or auxiliary prompt generators, leading to redundant computation. Although a single-stream alternative can reduce this cost, early fusion may also propagate noisy or misaligned auxiliary high-frequency cues through the backbone. In this paper, we propose a novel single-stream framework that integrates reliability-calibrated frequency adaptation into the adopted SAM backbone for MSOD. It avoids duplicated foundation backbones while explicitly controlling auxiliary frequency injection. Specifically, we design a mixture of frequency experts module, which uses the stationary wavelet transform to decompose each modality and aggregate cross-modal frequency information. We further introduce a reliability-calibrated frequency adapter with a dual-gate calibration mechanism, which selectively propagates the calibrated residual across transformer stages while jointly controlling its injection strength and cross-modal reliability. A hypernetwork-guided semantic-structural decoder then combines semantic mask features from the adopted backbone with Mamba-based structural detail recovery. Comprehensive experiments on RGB-D, RGB-T, and RGB-NIR salient object detection benchmarks validate that the proposed framework achieves competitive performance with only 12.20M trainable parameters, accounting for 5.4\% of the total parameters. The code will be available at https://github.com/xuboyue1999/SSSAM.

Recent advances in LLM agents have made them increasingly capable of designing methods for complex AI tasks. This raises two central questions about agent-designed methods relative to human-designed methods: how well they perform, and how different their algorithmic designs are. To study these questions, this paper introduces an analysis that derives task-specific algorithmic design spaces from human-designed methods, maps both human- and agent-designed methods into these spaces, and quantifies their algorithmic differences at the module level. Widely used LLM agents are evaluated on a suite of representative, open-ended AI tasks spanning multiple modalities, and the methods they design are analyzed in terms of both task performance and algorithmic differences from human-designed methods. Experimental results show that current agents can occasionally match or surpass human state-of-the-art (SOTA) performance (10/72 configurations), but such success does not generalize reliably across tasks or agents. Moreover, 96.8% of agent-designed methods fall within human-derived algorithmic design spaces, largely recombining algorithmic choices found in human-designed methods, while nearly half exactly match an existing human algorithmic design. Taken together, these findings suggest that although current agents can occasionally match or surpass human SOTA performance, their algorithmic designs remain within human-derived algorithmic design spaces, reflecting the reuse and recombination of algorithmic choices.

Storyboards turn screenplays into visual shot plans for automated short drama production. Professional storyboarding relies on tacit directorial expertise and remains an industrial bottleneck. Large language models can automate this step, but methods for supplying directing knowledge face three challenges: (1) Knowledge acquisition: the craft remains implicit in exemplars or must be written manually. (2) Knowledge refinement: authored knowledge is not evaluated against execution outcomes, and opaque generation prevents feedback attribution to the knowledge behind each decision. (3) Knowledge injection: injecting all knowledge exceeds usable context, while manual selection for every narrative group does not scale. We present SAGE (Skill with Attribution-Guided Evolution), a deployed framework that learns, attributes, evolves, and routes directing knowledge from expert demonstrations. SAGE derives rules that are independent of episode content by contrasting each training screenplay with its expert storyboard. During generation, the model records each narrative group's adopted rules. Combining these records with localized feedback enables targeted updates to individual rules. Evolved rules form scenario packages with a routing index, so each group retrieves only a bounded set appropriate to its situation without expert intervention. On 18 test episodes across three genres, SAGE scored 77.8 on a rubric validated by experts, versus 77.1 for professional directors. Deployed for 14 days on Virtual Film Studio, SAGE produced 1,344 narrative group outputs; 87.2 percent were accepted without substantive edits, and the production team recorded over 83 percent less authoring time per episode. We release PROSE, the first public dataset pairing screenplays with storyboards by professional directors across 68 episodes: https://github.com/creDreams/PROSE.

Regularized sparse regression has been extensively studied in the offline setting, but online formulation remains relatively under-explored. This gap stems from four key challenges: (i) the infeasibility of dynamically updating the regularization parameter in every online round, (ii) managing storage and memory complexity, (iii) enabling real-time computation via closed-form updates rather than solving full optimization problems at each round, and (iv) achieving optimal statistical guarantees under realistic assumptions. In this paper, we propose an online generalized-sparsity-constrained regression framework, focusing on online cardinality-constrained linear regression and low-rank matrix sensing. Unlike online regularized regression, our constrained formulation eliminates the need for dynamic parameter tuning. We introduce an efficient online hard-thresholding algorithm that performs closed-form updates and requires storing only summary statistics, making it computationally, memory, and storage efficient. Despite the inherent nonconvexity and combinatorial nature of the formulation, our algorithm achieves global convergence at the optimal statistical rate under realistic assumptions, provided that the projection set is properly overparameterized. Numerical experiments demonstrate that our method consistently outperforms state-of-the-art alternatives.

Long-horizon humanoid vision--language--action (VLA) control with head-mounted stereo cameras requires visual interfaces that can exploit complementary views while maintaining compatibility with pretrained representations. Existing interfaces often discard complementary stereo evidence or fuse additional observations without preserving the native primary-view pathway and adapting auxiliary information to robot embodiment. We present EATR-Stereo, an embodiment-aware token-routing framework that retains primary-view tokens and constructs primary-aligned Cross-View Auxiliary Tokens (CVATs) by querying the synchronized auxiliary-view token sequence. A body-segmented proprioceptive encoder further conditions token-wise auxiliary usage on robot configuration history, enabling selective incorporation of stereo evidence during action generation. The routed auxiliary stream augments the language and primary-visual context of a pretrained VLA while keeping its vision--language model frozen. On a 33-DoF physical humanoid with a 37-D proprioceptive state, we evaluate nine configurations in over-100-s search--approach--grasp--place--return tasks. EATR-Stereo achieves 60.0% full-task success, 100.0% grasp success, and 80.0% stage success. Under severe asymmetric occlusion, it improves recovery to 80% compared with 30% for CVAT alone. Ablation studies further show the importance of preserving primary tokens and combining cross-view auxiliary features with structured proprioceptive routing. These results demonstrate that selectively routed paired stereo evidence improves spatial grounding for reliable long-horizon humanoid VLA control.

Machine learning methods predict many real-world systems with remarkable accuracy, but they are typically treated as black boxes that offer no insight into which interactions drive the dynamics. Causal discovery methods reconstruct the interaction network from observational data, but without regard to whether the inferred structure supports prediction. Existing approaches combining both tasks rely on a single global hyperparameter, such as a causal threshold or a fixed neighborhood size, which cannot recover the structure of heterogeneous systems. Here we introduce causal local states (CLS), a framework that simultaneously infers an approximate Granger-causal interaction network and forecasts the system dynamics. For each node independently, we select the smallest set of neighbors that allows a predictive model to forecast the node near-optimally, and the resulting neighborhoods are then combined for a forecast of the full system. On three benchmarks of increasing difficulty, we achieve reconstruction of the underlying networks with high fidelity and forecasts on par with a model that is supplied with the true network, providing a step toward explainable and scalable forecasting of complex systems.

Knowledge Graph Reasoning (KGR) aims to discover latent facts by leveraging the structural evidence available in KGs, posing a challenge to the structural semantic understanding capability of KGR models. Recent studies have demonstrated that Large Language Models (LLMs) can achieve remarkable progress on KGR tasks via flexible in-context learning. However, the inherent representation inconsistency between KG structural context and LLM parametric knowledge remains inadequately addressed. This limitation prevents LLMs from effectively perceiving reasoning evidence that aligns with KG constraints, which undermines both the effectiveness and faithfulness of reasoning. We refer to this problem as reasoning evidence perception drift of LLMs over KGs. To address this problem, we propose a Structure-Internalized Rule Language Model (SIRLM), which centers on structural rule generation to couple the parametric learning of structural knowledge with the faithfulness evaluation of reasoning logic, enabling LLMs to anchor tightly to KG-grounded evidence. Specifically, we first design a Structure-Internalized Rule Generator (SIRG), which incorporates an in-context learning block augmented with a structural relation memory to coordinate structural and parametric knowledge. Furthermore, we equip SIRG with a KG tokenizer based on structural invariance learning and a neuro-symbolic reasoner based on rule-constrained message propagation. These components provide SIRG with learnable structural representations and faithful rule-execution feedback, respectively. Our SIRLM can be seamlessly integrated into standard LLM training paradigms, such as SFT and GRPO. Extensive experiments against 17 state-of-the-art KGR methods on 36 datasets demonstrate the significant superiority of SIRLM.

Although Graph Neural Networks (GNNs) have made significant advances in spatio-temporal traffic forecasting, their performance is limited when they rely solely on sensor proximity or road-network topology. This paper presents a spatio-temporal prediction framework, developed to incorporate knowledge in various forms. This framework aims to improve sensor-level, contextual understanding of the environment. A general-purpose knowledge graph (e.g., Wikidata) is used to create semantic subgraphs around traffic sensors and generate knowledge graph embeddings that capture meaningful relationships, such as nearby points of interest, administrative hierarchies, and the functional roles of locations. These embeddings are then fused with conventional traffic sensor graphs to provide additional adjacency matrices informed by semantics. This allows GNNs to learn the semantic context beyond physical connectivity. This study differs from previous research in two key ways. Firstly, rather than proposing a novel GNN architecture, it demonstrates the general impact of external knowledge on prediction accuracy. Secondly, experiments with well-established traffic forecasting approaches show that external knowledge provides additional information that street network data alone cannot convey. The results show that integrating data from general-purpose knowledge graphs and sensor networks through data fusion can enhance the prediction accuracy of traffic forecasting models, and offers a potential pathway toward improved interpretability.

LLM agents have been widely adopted to operate mission-critical infrastructure (MCI). These agents normally rely on a harness that determines what information they can access, which tools they can use, and what actions they can take. Existing systems often expose the same comprehensive harness to every task, which may not be necessary and cause resource wastes. In this paper, we focus on the identification of optimal harness configurations, and view it as a resource-matching problem between what each task requires and what the harness provides. To measure this match, we classify MCI tasks based on the mathematical representation of the underlying system and rank harness configurations by the amount and type of information they provide. We then construct task-to-harness mappings from two sources: mining research literature and measuring controlled agent execution. Leveraging the measured mapping, we propose a new harness provisioning algorithm: map-guided escalation. It begins with a task-specific harness and expands to full provision only after a failed self-check. We evaluate our method in two representative MCI tasks: in liquid cooling, it improves the agent accuracy from 0.652 under full provision to 0.715 and achieves accuracy comparable to Reflexion with 48% fewer tokens; In power grids, full provision remains accuracy-optimal, while map-based provisioning offers lower-cost alternatives. These findings show that harness provisioning follows a domain-dependent accuracy-cost Pareto frontier rather than a universal optimum.

Generative imitation learning policies excel at trajectory planning but lack closed-loop force regulation, while directly incorporating force modalities often requires redesigning or retraining the network. We present UniReflex, a universal plug-and-play framework that equips frozen generative policies with variable impedance control (VIC) for contact regulation, guided by force-direction intent collected during demonstration, without further slow-backbone fine-tuning. By non-invasively intercepting deep latent representations from the action head, UniReflex drives a fast reflex network that decouples active force exertion from external interaction response. This scheme predicts normalized anisotropic stiffness directions for directional compliance allocation. Furthermore, UniReflex integrates an adaptive gating mechanism that enables seamless transitions between position-dominant planning and force-dominant execution. Real-world bimanual experiments demonstrate that UniReflex significantly improves contact stability and success rates while preserving original position accuracy. Our approach achieves 25-66x lower per-step backward latency relative to joint training strategies on the evaluated backbones.

We introduce Semantic Task Completion Video Generation, an outcome-oriented video generation task. Under this formulation, success requires both achievement of the intended outcome and semantic grounding. Semantic grounding characterizes the correspondence between the reference image and the generated outcome in terms of high-level semantics relevant to the task. Evaluation focuses on the generated outcome and requires neither the presentation of a complete sequence of intermediate task steps nor conventional appearance consistency with the reference image. To support systematic evaluation, we construct SemComp-Data, an evaluation dataset covering six domains. Each instance comprises a reference image, a detailed instruction, a brief instruction, and an outcome-centric video clip. A scalable four-stage curation pipeline converts raw videos into standardized SemComp-Data instances. We further introduce SemComp-Bench, an evaluation protocol that uses a vision-language model (VLM) to answer structured binary questions. SemComp-Bench reports the OA Score and the GR Score for Outcome Achievement and Generation Reliability, respectively. Experiments on representative video generation models show that achieving intended outcomes while maintaining task-relevant semantic grounding in reference images remains challenging.

Effective segmentation of multi-modal MRI is central to improving neural network accuracy in brain tumor recognition. Existing methods typically compress 3D volumes into token sequences via fixed patch encoding or learned attention pooling (e.g., TokenLearner). However, these compression schemes discard explicit spatial shape information; the resulting tokens convey no notion of lesion morphology or spatial extent. Meanwhile, end-to-end evaluation entangles a tokenizer's information retention with the reconstruction capacity of the downstream decoder, and the lack of a unified capacity contract across methods makes performance differences difficult to attribute. In this paper, we introduce Gaussian tokens to multi-modal brain tumor segmentation for the first time: each token carries not only a semantic feature but also a learned 3D center, anisotropic scale, and orientation, endowing the representation with explicit geometric support at negligible parameter cost. We further propose a frozen-token utility evaluation protocol: the trained tokenizer is frozen, its output is cast into a fixed-capacity serialized contract, and a shared lightweight Transformer probe independently measures each tokenizer's retained information under strictly matched conditions. Multi-seed paired statistical testing shows that GSToken consistently and substantially outperforms capacity-matched adaptive baselines under frozen probing, with uniform advantages across all tumor sub-regions, surface, and distance metrics. These results demonstrate that explicitly encoding spatial geometry within tokens significantly improves the information density of volumetric representations, offering a new design principle for compact 3D medical image representation and downstream reading.

GRPO is increasingly used for reinforcement learning of vision-language-action (VLA) policies because, unlike PPO, it does not require training a critic. This simplification comes with a sampling cost: group-relative advantages require multiple rollouts from each scene. Under binary success rewards, groups whose rollouts all succeed or all fail have zero advantage and are discarded by dynamic sampling. These groups are especially common early in training, when most rollouts fail, wasting much of the expensive robotic rollout budget. We introduce Prism-GRPO, which augments binary outcome reward with a weighted trajectory-level execution-quality score. By splitting same-outcome groups into a quality spectrum, Prism-GRPO recovers training signal while ensuring that every success still outranks every failure. Quality scores can be derived from simulator contacts, executed actions, or visual observations, avoiding task-specific progress rewards. We prove that Prism-GRPO never increases the probability that a sampled group is discarded for having zero advantages, and derive a gradient-alignment condition under which its combined update remains a local ascent direction for task success. Across four RoboTwin tasks spanning different horizons and coordination patterns, Prism-GRPO improves success and quality at matched rollout budgets and reaches target success rates with up to 56% fewer rollouts. It also suppresses a reward-hacking shortcut, with the cleaner behavior transferring under direct deployment to a real robot. Through ablations, we show consistent gains across contact-, smoothness-, and VLM-derived quality signals.

Differentially private training adds isotropic Gaussian noise to clipped gradients, corrupting every singular direction equally. In vision models, where spatial correlation concentrates gradient energy into a low-rank subspace, most of this noise falls in directions that carry little signal. Spectral gradient orthogonalization via polar decomposition is introduced as a post-processing step that recovers directional signal from the noisy gradient's low-rank structure at zero additional privacy cost. A phase transition governs the utility of this approach: orthogonalization improves accuracy only when the per-direction spectral signal-to-noise ratio (SNR) suffices for singular vector recovery; in low-SNR regimes, the directional bias of the gradient is replaced by a nearly random orthogonal update, and the transformation is harmful. The recovery threshold is determined by the spectral gap of the gradient and is surpassed at large batch sizes. Empirically, the benefit scales with model capacity: spectral orthogonalization achieves a +20.9% improvement over DP-SGD on WRN-28-10 (B = 4096) and +14.9% on ResNet-18, while reducing inter-run variance by a factor of two to three. In the fine-tuning regime, spectral orthogonalization matches the stability of DP-Adam while maintaining a first-order memory footprint. Combining spectral with temporal denoising yields 50.3% on CIFAR-10 (epsilon = 4), the highest accuracy in any tested configuration. These gains are specific to moderate-to-high-SNR regimes such as large-batch training of higher-capacity models. Small-batch or low-SNR settings are better served by DP-SGD or temporal denoising.

Chart editing requires inferring and modifying visualization code from a reference chart image based on an editing instruction, challenging fine-grained visual reasoning, instruction following, and executable code synthesis capabilities of MLLMs. Large reasoning models (LRMs) with extended Chain-of-Thought (CoT) reasoning are suitable for tackling such complex multimodal tasks. However, our preliminary study reveals an ``inverted-U'' relationship between reasoning length and chart-editing performance: Excessive reasoning often leads to ``overthinking,'' where models drift toward hallucinated visual details or get stuck in redundant reasoning loops. To address the gap, we introduce REChart, a two-stage training framework that provides process-level supervision over intermediate reasoning steps, improving both editing fidelity and reasoning efficiency. First, we synthesize 200k high-quality reasoning trajectories for supervised fine-tuning from a large image-instruction-code pool, using a role-specialized agentic Reason-Score-Refine workflow that iteratively refine the chart code toward higher quality. Second, we optimize the model via reinforcement learning with two complementary rewards: a \emph{fidelity} reward evaluating code correctness, visual fidelity, and structural consistency, and an \emph{efficiency} reward that assigns each rollout a random thinking budget, truncates the reasoning process, and credits the final reasoning segment according to its contribution to the output. On the ChartEdit and ChartMIMIC benchmarks, our model achieves state-of-the-art chart-editing performance among open-source models of comparable scale, while mitigating overthinking and reducing average reasoning token usage by 79.0\% under a maximum thinking budget of 16,384 tokens compared with the base model.

Group Relative Policy Optimization (GRPO) has become a widely used approach for post-training Large Language Models (LLMs) for reasoning. In GRPO, the group gradients induced by different queries within the same mini-batch are directly averaged to form the policy update. However, these group gradients can point in conflicting directions. Our empirical analysis suggests that group-gradient conflicts tend to be associated with less effective policy updates, motivating the need for a reliable aggregated update direction under such conflicts. Standard GRPO aggregation treats the realized group gradients as deterministic contributions and does not account for differences in their reliability during aggregation. To address this issue, we propose Gradient Uncertainty-Aware Policy Optimization (GUPO), which models each group gradient as a random variable under a Bayesian formulation and estimates its probability distribution. GUPO then derives gradient uncertainty using a Dirichlet-based formulation and uses it to calibrate the contribution of each group gradient during aggregation. Extensive experiments on multiple benchmarks demonstrate the effectiveness of GUPO.

Vision encoders are a critical component of vision-language models, and scaling their capacity effectively improves performance. However, dense scaling increases compute cost and inference latency. Mixture-of-Experts (MoE) architectures offer a compelling alternative, having enabled efficient scaling in LLMs, yet the MoE design space for CLIP-style vision encoders remains underexplored at State-of-the-Art (SOTA) levels. In this work, we systematically study MoE designs for vision encoder scaling and find that fine-grained MoE topologies yield substantial gains over both dense and standard MoE counterparts. We further propose an auxiliary-loss-free balancing variant for better expert utilization, and design a specialized MoE kernel to mitigate inference latency overhead. To enhance video capabilities while preserving image knowledge, we introduce frame-level distillation paired with a novel freezing mechanism. We pretrain a series of Mixture-of-Experts Vision Encoders (MoE-ViE) across a range of sizes, all consistently outperforming their dense counterparts. Our largest model matches the zero-shot performance of a SOTA encoder 1.7x its size at 76% of its latency. When aligned with an LLM, MoE-ViE surpasses all compared encoders on image and video benchmarks, including those with up to 5x more activated parameters. Code is available at https://github.com/facebookresearch/moe_vie.

Reinforcement learning for coding agents increasingly relies on long-running agent harnesses to manage tool integration, repository contexts, and execution feedback. However, the native execution environments of these harnesses are inherently misaligned with policy-gradient training: environmental crashes and reward hacking corrupt outcome signals, while train-inference discrepancies decouple rollout behavior from policy updates. To address this, we present LEGO-RL, a framework that bridges native coding-agent harnesses with scalable policy-gradient optimization without modifying their internal control flow. LEGO-RL is built upon three pillars: (1) faithful optimization via in-process LLM proxying that captures raw generation streams for token-level alignment and robust trainer-side log-probability recomputation, even under harness-side compaction or re-serialization; (2) reliable execution via scalable sandbox orchestration featuring image caching and stage-wise defenses to mitigate reward hacking; and (3) observable training through an integrated plugin that automates validation and monitoring, paired with a Live UI for granular trajectory diagnostics. We evaluate LEGO-RL by training the sparse MoE model Qwen3.5-35B-A3B with GSPO across three native coding-agent harnesses. LEGO-RL improves Qwen3.5-35B-A3B across OpenHands SDK (64.0% to 70.4%), Claude Code (62.4% to 68.2%), and OpenCode (57.2% to 66.6%) on SWE-bench Verified, while maintaining a rollout-training probability correlation above 0.99.

Long-sequence 3D reconstruction from RGB videos requires both accurate local geometry and globally consistent camera motion. Feed-forward models provide strong depth and pose predictions, but their memory cost prevents joint inference over long sequences. Chunk-wise processing improves scalability, yet independently predicted chunks often exhibit scale drift, pose errors, and point-cloud misalignment. We present GeoWeaver, a unified framework comprising a Geometric Prior Model (GPM) and Test-Time Adaptation (TTA). The GPM predicts chunk-wise depth, confidence, and camera parameters as adjustable geometric priors. TTA then performs sequential initialization, global chunk-level Sim(3) alignment, and coarse-to-fine refinement of camera poses, affine depth corrections, and intrinsics. Dense correspondences provide adjacent, cross-chunk, and long-range constraints, while a robust CDF-style objective jointly optimizes weighted 2D reprojection and 3D consistency residuals. This design preserves local geometric accuracy while correcting accumulated pose, scale, depth, and calibration errors. Experiments across diverse long-sequence benchmarks demonstrate improved camera accuracy, global consistency, and point-cloud quality. Ablations verify the contribution of each adaptation stage, and applying the same TTA procedure to different geometric prior models consistently improves their trajectory estimates, demonstrating that GeoWeaver is not tied to a specific GPM.

Foundation-model policies for robotic manipulation are advancing rapidly on task success, but rigorous evaluation of whether they succeed safely is still lacking. We introduce ManiGuard, a specification-grounded framework for evaluating and improving the safety of foundation-model manipulation, comprising the ManiGuard-Bench task suite and a paired safety-annotated trajectory-generation pipeline. ManiGuard-Bench organizes six contact-rich household task families into 200 locked base tasks along a skill $\times$ constraint taxonomy, with safety specified independently of task success. Each task is evaluated under one in-distribution and four single-axis out-of-distribution perturbations that hold the safety specification fixed, giving 1,000 locked scenarios. Every rollout is runtime-checked by LTL$_f$-grounded automaton monitors over physics-grounded predicates rather than learned classifiers or LLM judges, in simulation and on a physical Franka platform. The pipeline pairs an automated motion-planning generator with human teleoperation, annotated by the same per-step monitor, and directly supports safety-aware fine-tuning; we release 8,000 safety-annotated demonstrations, 40 per base task. Benchmarking zero-shot and fine-tuned VLAs across more than 23,000 rollouts, we find: (i) safety must be evaluated independently of task success, as 6-21% of successful rollouts violate the specification; (ii) fine-tuning on our suite raises safe task completion from near zero to 7.5-29.8% and engaged-and-safe behavior from 16-40% to 51-72%; but (iii) a gap remains that scaling demonstrations does not close, with 21-42% of engaged rollouts still violating, two of six families below 2% safe success for every policy, and these failures persisting under distribution shift and on hardware.

We introduce PTXBench, a benchmark for evaluating and adapting large language models (LLMs) to use architecture-specific PTX for GPU kernel optimization. PTXBench measures functional correctness, whether selected target instructions execute at runtime, and speedup over frontier libraries across GEMM and attention workloads on H100 and B200 GPUs. Our evaluation shows that architecture-specific PTX capability remains uneven: success rates fall substantially on complex attention backward workloads, and executing the target instructions does not necessarily translate into competitive performance. No evaluated model consistently matches frontier libraries across the suite. We further adapt Qwen3.6-27B using supervised fine-tuning. Repair-conditioned training improves several tasks, but generalization remains uneven; data coverage, balance, and the quality of the reasoning teacher matter in addition to dataset size. PTXBench provides an auditable testbed for measuring and improving LLMs' ability to exploit evolving GPU architectures.

Generative models promise a route to explainable clinical AI: rather than probe a classifier, model the distributions of healthy and diseased patients and read explanations off the geometry between them. We build such a system - an optimal-transport rectified flow trained between two clinical distributions - and use it to ask a pointed question the field too rarely tests: do the resulting explanation heatmaps actually localize disease? On tabular tumour biomarkers (Breast Cancer Wisconsin) a single flow yields per-patient counterfactuals, an unsupervised malignancy score (AUROC 0.91; 0.93 +/- 0.01 across five seeds), and a label-free attribution that agrees with a supervised classifier (r ~ 0.5) - a compact, honest interpretability engine, though it never out-predicts logistic regression. Moving to chest X-rays, we show the transport heatmap is a population-level signal, not a localiser; a reconstruction-based, identity-preserving variant does localize synthetic lesions (pointing game 0.52), yet on real RSNA radiologist boxes it collapses to chance while only supervised Grad-CAM stays above it. The central result is a synthetic-to-real gap: label-free heatmaps that look compelling on planted lesions are not evidence of real localisation. We contribute a reusable optimal-transport recipe for generative explanations and a controlled benchmark for stress-testing whether they localize.

In this work, we demonstrate a data-driven machine learning (ML) framework for the efficient design and optimization of Kretschmann-configuration-based surface plasmon resonance (SPR) sensors. A physics-based dataset was generated using a MATLAB-based transfer matrix method (TMM), covering diverse material properties, layer thicknesses, and multilayer configurations. Optical properties and layer thicknesses were used as input features, while figure of merit (FOM) and minimum reflectance (Rmin) were the target performance parameters. Four ML models, namely CatBoost, XGBoost, LightGBM, and multilayer perceptron (MLP), were benchmarked using R2, mean absolute error (MAE), and root mean squared error (RMSE). The framework integrates ML benchmarking, SHAP explainability, robustness analysis, and optimization-driven inverse design. SHAP-weighted perturbation experiments assessed the model's robustness to input variations. Four optimization algorithms were employed for inverse sensor design, followed by an analysis of parameter recovery and performance. The optimizers were also evaluated using an independent forward-design task, in which repeated runs converged on a common configuration. The optimized designs agreed closely with direct TMM calculations, with FOM errors of 0.6-0.9 percent and Rmin errors below 0.7 percent. The ML models achieved R2 values greater than 0.99 while reducing computational cost from seconds to milliseconds, corresponding to an acceleration of approximately 10^3 to 10^4 times compared with direct TMM simulations. Overall, the results demonstrate that physics-based surrogate ML models combined with explainability and optimization provide a computationally efficient and interpretable framework for rapid SPR sensor analysis, inverse design, and optimization.

Industrial defect detection differs from natural-image object detection because inspection images are captured under controlled conditions and contain large normal-dominant regions with repetitive structures. Defects therefore appear as localized disruptions of otherwise predictable patterns, while conventional detectors rely mainly on sparse bounding-box supervision, resulting in weakly constrained normal-region representations. We propose a continuity-driven representation regularization framework that exploits normal-dominant regions as dense auxiliary supervision. The framework introduces two detector-agnostic objectives: Multi-Continuity Loss, which combines 1D patch-sequence prediction and 2D masked spatial prediction, and Differencing Loss, which regularizes first-order feature variation and second-order curvature between neighboring patch embeddings. Both objectives are applied with box-derived region weighting to stabilize normal-region representations while preserving defect-related discontinuities. Experiments on two real-world industrial datasets and the public NEU-DET benchmark, using six detector architectures including YOLO-family models, MambaYOLO, and DETR, demonstrate consistent improvements over native detector baselines. In the full-data setting, the proposed regularizers improve average mAP@0.5:0.95 by up to 3.49 percentage points on Industrial Metal, 5.38 percentage points on MEA, and 5.03 percentage points on NEU-DET. Under limited-data conditions, the gains become more pronounced, with Differencing Loss achieving improvements of up to 21.07 percentage points in mAP@0.5 and 8.23 percentage points in mAP@0.5:0.95 on NEU-DET using only 25% of the training data. These results suggest that continuity-driven regularization provides an effective prior for improving industrial defect detection, particularly when annotated data are scarce.

Reliable jailbreak evaluation is essential for assessing LLM safety, but most existing studies rely solely on attack success rate (ASR) without accounting for its dependence on attack budgets, resulting in unfair comparisons across methods. Existing compute-aware evaluations reduce heterogeneous resources into FLOPs, which is difficult to estimate for black-box models and fails to capture resource-specific constraints. To provide a comparable evaluation basis, we introduce Fair-ASR, an evaluation protocol for black-box jailbreak attacks under shared target-call budgets B, using target calls as a directly observable and method-agnostic comparison axis while tracking attacker calls separately for efficiency analysis. We re-evaluate 11 representative attacks under the Fair-ASR protocol and find that attack rankings change substantially across target-call budgets, simple stochastic perturbations and hand-crafted templates remain highly competitive under equal target access, and no evaluated LLM-driven method is efficient in both target and attacker calls. Motivated by this efficiency gap, we introduce ReCode, a compositional budget-efficient attack that combines desensitization rewriting with two effective low-cost primitives identified by Fair-ASR. Under a budget of 20 target calls, ReCode achieves 85% ASR on GPT-5 while requiring only 7.19 attacker calls per request on average, showing strong efficiency in both target and attacker calls.

Most automated essay scoring (AES) systems output a single holistic score without interpretable evidence and rely on closed APIs that introduce data privacy and cost barriers. We present ArguLens, an opensource, locally deployable system that decomposes AES into three decoupled components: a discourse-move classifier (Qwen2.5-7B-Instruct fine-tuned with LoRA on PERSUADE 2.0), a grade-independent LightGBM scorer over 31 linguistic and discourse features, and a label-aware feedback generator served through vLLM with a Qwen2.5-14BInstruct backbone. A Gradio web UI exposes pluggable inference backends and supports single-essay and batch scoring with downloadable per-essay breakdowns. On an essaydisjoint PERSUADE 2.0 test split, the logitprobe classifier achieves 82.6% accuracy and 0.727 macro-F1; under prompt-grouped 5-fold cross-validation the scorer reaches a mean QWK of 0.813 under an oracle discoursefeature protocol, and an ablation shows that adding gold discourse annotations yields an increment of +0.055 QWK over the lexical+syntactic configuration (paired t-test, p = 0.010). This is a component-level diagnostic rather than an end-to-end classifier-to-scorer result. The feedback generator ships with a structured evaluation protocol; its human-rater study is left to future work. The system is released under Apache 2.0 at https://github.com/wwrwbs/AI_AWE.

Distributed Denial-of-Service (DDoS) attacks threaten network availability, requiring a cognitive detection process that senses traffic, infers intent, and supports an adaptive response under severe class imbalance and non-stationary conditions. This paper proposes a Graph-based Generative Adversarial Network (GraphGAN) that serves as the cognitive detection engine for this task. GraphGAN captures the relational structure among traffic flows while addressing imbalance through adversarial generation of synthetic samples. Sequential flows are converted into $k$-nearest neighbor graphs using sliding windows to preserve feature-similarity and temporal dependencies among flows. The generator learns the distribution of DDoS attacks to synthesize realistic minority samples, while a Graph Convolutional Network (GCN)-based discriminator distinguishes real from synthetic graph data. A separate GCN classifier, trained on the balanced dataset, performs the final detection decision. Evaluations on four benchmark datasets show that GraphGAN achieves superior accuracy, precision, and recall compared to state-of-the-art approaches, particularly in data-scarce scenarios. By integrating temporal graph construction, adversarial augmentation, and GCN classification, GraphGAN effectively models coordinated attack behaviors and mitigates class imbalance, providing a robust and topology-aware solution for intrusion detection in data-constrained environments.

Repetition is a fundamental mechanism in human learning, where revisiting successful experiences strengthens memory, consolidates skills, and improves future performance. Motivated by this biological principle, we introduce Instant Episode Repetition (IER), a simple and novel mechanism that improves sample efficiency by immediately repeating action sequences from successful episodes during environment interaction. Unlike conventional approaches such as Experience Replay and Self-Imitation Learning (SIL), which passively reuse past experience during training updates, IER directly influences the data collection process. Upon identifying a high-reward episode, the agent repeats its action sequence for a fixed number of subsequent episodes, reinforcing valuable behaviors through renewed interaction with the environment. We integrate IER into state-of-the-art SAC and TD3 algorithms and evaluate its effectiveness on continuous-control benchmarks, including MuJoCo, the DeepMind Control Suite, and a real-world dynamic object translation task with a robotic manipulator. Experimental results demonstrate that this simple mechanism improves learning performance over standard and self-imitation-based baselines.

Neural network surrogates are an emerging alternative to traditional electromagnetic wave simulators like finite-difference time-domain (FDTD); their goal is to replace rigorous physical simulations with pre-trained neural networks that solve wave-scattering forward and inverse problems orders of magnitude faster. However, nonrecurrent, single-step surrogates have scaled only to a few tens of simulation variables. Here, we show that this barrier can be overcome by dynamically generating salient training examples during training, rather than randomly sampling the large space of possible examples. We introduce an algorithm that runs in parallel with surrogate training, using gradient ascent to search refractive-index and source configurations for cases where the surrogate disagrees with a full-wave ground-truth simulator. We also use source and ground-truth normalization with an evolving replay dataset to stabilize and accelerate learning. Using this approach, we train a fast, single-step surrogate for two-dimensional wave scattering with up to 41,772 controllable variables, including dense, freely configurable grids of refractive indices and complex-valued sources. The resulting neural surrogate is robustly accurate across diverse structured and unstructured examples and generalizes inductively to larger domains, reaching over 3 million controllable variables without retraining, a $73.8\times$ increase. We demonstrate the surrogate on large-scale forward simulations and inverse design of freeform beam splitters and gradient-index (GRIN) lenses up to 98 wavelengths wide, showing comparable or better performance than FDTD-based designs, with speedups from $1.29\times$ to $26.5\times$. These results demonstrate a practical path toward fast, robustly accurate, inductively scalable neural simulators for photonic inverse design and other wave-scattering inverse problems.

Long-context prefill in large language models (LLMs) incurs substantial computation and memory traffic because dense self-attention computes quadratic query-key scores. Existing methods either use a uniform low-precision path or select token interactions, leaving spatial precision routing over hardware-aligned score tiles outside fused dense attention. We introduce TileMix, a tile-centric precision-routing kernel that makes numerical precision an executable spatial decision over score-tile groups within fused dense attention. TileMix partitions the attention matrix into hardware-aligned score tiles, packs routing decisions into compact bitmasks, and dispatches each tile group through FP16 or INT8 score computation while both paths update a shared online-softmax state. Scalable precision grouping lets each routing bit govern multiple adjacent key tiles, preserving hardware-aligned compute tiles and compact metadata at long contexts. By routing all legal tile groups, TileMix preserves dense token connectivity, requires no training, and supports grouped-query attention, variable-length batches, and INT8 key/value caches. Across LongEval, LV-Eval, and A100 prefill benchmarks on LLaMA, Qwen, and Vicuna, TileMix recovers long-context quality lost under uniform INT8 and improves prefill throughput over FP16, yielding a controllable accuracy-efficiency frontier across model families. The implementation is available at https://github.com/HanzhiZhang-Ulrica/TileMix.

Facial biometric recognition systems currently face compound threats intertwining generative AI and high-fidelity physical spoofing. Existing defenses suffer from systemic bottlenecks, including poor generalization, non-auditable reasoning, and reliance on massive, low-quality datasets. To address these challenges, we propose Multimodal Large Language Models (MFAD) for face anti-spoofing detection, an explainable reasoning system for Unified Face Anti-Spoofing Detection (UFAD), accompanied by a semantic-level annotation benchmark. Unlike methods relying on external tools or coarse alignment, MFAD activates the intrinsic reasoning capabilities of Multimodal Large Language Models (MLLMs) via a fine-grained pixel-semantic anchoring mechanism. This eliminates localization hallucinations and ensures auditable reasoning paths. We introduce a cross-attack semantic-level unified annotation paradigm: by annotating only 1,000 precise masks per attack category, we generate reasoning evidence chains strictly corresponding to spoofed regions. Supervised fine-tuning on the Qwen-VL foundation model demonstrates that, using limited high-quality samples, the system achieves a 40-50% relative reduction in in-domain ACER and restricts cross-domain performance degradation to within 11.62%/5.23%, significantly outperforming existing frameworks. Furthermore, under white-box adversarial attacks, detection accuracy drops by only 3.2%, validating the robustness of semantic anchoring compared to models trained on massive short-text data. Domain practitioners rated the evidence reliability of reasoning paths at 4.57/5, with inference latency satisfying real-time deployment requirements. These results confirm that a few-shot, high-quality semantic annotation paradigm is effective for building trustworthy, explainable, and cost-efficient UFAD systems.

Robotic manipulation policies trained via imitation learning, such as Action Chunking with Transformers (ACT), can achieve strong performance under ideal conditions but often remain sensitive to small execution errors and distribution shifts. Correcting these failures typically requires dataset aggregation and full-policy retraining, which is computationally expensive and unsuitable for real-time deployment. In this work, we propose Online Residual Policy Adaptation (ORPA), a framework that enables immediate, feedback-driven correction of robot actions without modifying the underlying policy parameters. ORPA augments a pretrained control policy with a lightweight, feedback-conditioned module that predicts residual adjustments directly in joint space, allowing the system to adapt its behavior at runtime. We evaluate ORPA on a set of precision-sensitive manipulation tasks using the ALOHA platform, demonstrating improvements in success rate and recovery from small perturbations compared to baseline control policies and rule-based inverse kinematics corrections.

Brachiation is a form of locomotion in which primates move primarily using their arms, enabling traversal in environments without footholds. However, this motion requires highly coordinated whole-body movement and precise timing control for bar grasping and release. As a result, achieving robust behavior on life-sized robotic platforms remains challenging. In this study, we present a reinforcement learning-based method to realize brachiation on a life-sized dual-arm robot. The core of the proposed approach is Waypoint-Guided Reinforcement Learning (WGRL), a learning framework for inducing non-linear and complex motions. For high-difficulty tasks where imitation learning data are unavailable, WGRL guides behavior acquisition by sparsely specifying waypoints for the end-effector trajectory, while whole-body motion is generated through reinforcement learning. In addition, by integrating the waypoint-following guidance with rewards based on task success and mechanical energy, and training in an environment designed for Sim-to-Real transfer, the proposed method achieves both forward progression and motion stability. The acquired behavior is evaluated through Sim-to-Sim experiments under monkey-bar environments with geometric variations and hardware experiments, confirming robust brachiation including failure recovery behavior. This study provides effective learning design guidelines for realizing arm-based locomotion on life-sized robotic hardware and expanding the traversable workspace of robots.

Browser agents perform well on short, clean demonstrations, but real deployment is fundamentally different: agents must sustain dozens of decisions on live websites while recovering from mistakes and navigating complex UIs. We argue that closing this gap requires alignment at every level of the pipeline, including execution, supervision, optimization, and evaluation, rather than scale alone. We present Wuying-Browser-Agent, a unified framework that addresses each of these levels. A structured browser harness provides stable execution primitives and decision-oriented context management. Reflection and UI-specialized Curriculum SFT (RUIC-SFT) explicitly trains on recovery trajectories and complex-UI interactions. Divergence-Aware Online GRPO (DAO-GRPO) improves long-horizon credit assignment through potential-based reward shaping and divergence-aware step weighting. Finally, we introduce BrowserBench, a bilingual real-web benchmark of 350 tasks averaging 37.9 steps, because most existing benchmarks are too short to expose long-horizon failure modes. Wuying-Browser-Agent-27B achieves 80.6\% on WebVoyager, 66.7\% on Online-Mind2Web, and 65.1\% on BrowserBench, establishing a new open-source state of the art on browser-use benchmarks. The same pipeline also transfers beyond browser use, demonstrating strong general agentic ability and reaching an average score of 73.8 on Tau2-Bench, Claw-Eval, and BFCL-v4.

Vision-language navigation agents are often evaluated on their ability to follow route-like instructions toward a fixed goal. Yet, real navigation instructions often depend on observed states of the environment: if a condition holds, then follow one path, otherwise take another. Such instructions require an agent to evaluate scene evidence, select the correct logical branch, and execute the corresponding navigation behavior. Existing evaluations provide limited control over conditional branch execution, making it difficult to determine whether agents fail because of perception, grounding, navigation, or logical decision-making. We introduce CondVLN, a scene-graph-grounded benchmark for diagnosing conditional branching in vision-language navigation. CondVLN programmatically generates instructions whose branch conditions are grounded in verifiable 3D scene-graph predicates, with controlled variation in branch depth, dependency chain length, spatial composition, evidence observability, and instruction horizon. CondVLN contains over 11,500 generated conditional instructions across AI2-THOR, Matterport3D, Gibson, and ReplicaCAD, and evaluates agents using standard VLN metrics and branch-specific diagnostics: Branch Selection Accuracy and Conditional Success Rate. Evaluating four state-of-the-art VLN agents (VLN-Zero, NaVid, NaVILA, and Open-Nav) shows that conditional branching exposes failures that are not captured by standard success rate or path length alone: agents can navigate plausibly while committing to a branch inconsistent with the observed scene condition. We also present a lightweight neurosymbolic branch-selection model that separates condition grounding from navigation execution, improving performance by 2x. CondVLN provides a reusable testbed for measuring whether embodied agents can not only follow instructions, but follow the right instruction under the right condition.

Animatable human avatars are routinely reconstructed from multi-view video under a silent assumption: that every pixel of a frame observes the same instant of the body's motion. Rolling-shutter (RS) sensors expose image rows sequentially, so within one frame the head and the feet of a moving person are separated by tens of milliseconds of articulated motion, and every scanline sees a different pose. Feeding such video to a state-of-the-art avatar bakes the distortion into the canonical representation, where it survives as shear and wobble under novel views and novel poses. Worse, every camera in a rig follows its own readout schedule, so the multi-view consistency that drives the reconstruction is violated even when the geometry is correct. We present RS-Avatar, which reconstructs a sharp, undistorted, animatable 3D Gaussian avatar directly from RS video. The formulation is minimal: a motion-aware avatar already renders the body at several sub-frame instants, and where a blur model averages those renderings, a rolling-shutter model composites them scanline by scanline. Changing that operator is the only modification required. On RS-ZJU, a benchmark we build from ZJU-MoCap, this improves novel-view synthesis over training as if the frames were instantaneous, on every subject. A motion-aware blur model built on the same sub-frame machinery does not transfer, and in fact falls below the shutter-oblivious baseline: the machinery is reusable, the operator is not.

Reinforcement Learning (RL) has been promising in single-turn LLM fine-tuning. However, long-horizon agentic reasoning introduces increasingly branching interactions and sparse rewards, exposing several limitations of RL: its heavyweight backpropagation-based training stack makes it impractical to fine-tune larger LLMs, and longer-horizon trajectories make credit assignment in RL substantially harder. This paper argues that evolution strategies (ES) can be a better choice for fine-tuning long-horizon LLM agents. Compared with agentic RL, ES offers three key advantages: 1) Model Scalability: ES enables full-parameter optimization with only minimal, inference-level GPU memory, making it possible to fine-tune large LLMs. 2) Flexibility: its lightweight, black-box feedback interface makes ES fine-tuning easy to compose with prompt-space evolution (e.g., skill optimization & test-time compute); and 3) Long-Horizon Scalability: ES performs trajectory-level parameter attribution without decomposing rewards across horizons, yielding better scalability than Agentic RL as the horizon length grows. Based on this insight, we propose Agentic ESOpt, a full-parameter agentic fine-tuning framework tailored to flexible parameter--context co-evolution. At each step, Agentic ESOpt samples perturbations around the current LLM parameters, evaluates the resulting agents with rewards, and applies an online reward-weighted update. To improve the exploration--adaptation trade-off, Agentic ESOpt further introduces a cosine decay schedule of the perturbation scale $σ$. On WebArena-Lite, full-parameter optimization of Qwen-3.5-27B improves the No Skill baseline by 6.69%. In test-time automatic heuristic design, Agentic ESOpt performs online prompt--parameter co-evolution, improving its matched baseline in 28 of 36 settings.

Post-training with supervised chain-of-thought fine-tuning and reinforcement learning from verifiable rewards has substantially improved the mathematical reasoning capabilities of large language models (LLMs). However, their application to signal processing problems remains relatively under-explored. This report investigates reinforcement fine-tuning strategies for adapting Qwen2.5-3B-Base to graduate-level signal mathematical problems from WirelessMATHBench-XL, a comprehensive benchmark for mathematical reasoning in this domain. We examine two training paradigms: (i) direct reinforcement learning (RL) on WirelessMATHBench-XL with verifiable rewards; and (ii) supervised fine-tuning (SFT) on a distilled wireless-domain chain-of-thought corpus, followed by the same domain-specific RL stage. Across both paradigms, we benchmark Group Relative Policy Optimization (GRPO), Group Sequence Policy Optimization (GSPO), and Geometric-Mean Policy Optimization (GMPO). We aim to assess whether domain-aware CoT SFT serves as an effective initialization for subsequent RL, and whether GSPO or GMPO offer advantages in stability or accuracy over GRPO for signal reasoning tasks. Our best model achieves an overall accuracy of 39.12\%, representing a more than threefold improvement over the untrained Base model (12.37\%).

Time Series Foundation Models (TSFMs) have recently emerged as a highly promising paradigm for cross-domain zero-shot forecasting. However, existing evaluation protocols predominantly rely on static benchmarks with fixed historical test windows. While these benchmarks provide a valuable baseline snapshot, they evaluate an average performance on a fixed history, failing to capture how models behave in continuously evolving real-world environments characterized by seasonal variations, distribution shifts, and unexpected events. To bridge this gap, we introduce LiveHouse-TS, the first open-world living benchmark infrastructure for TSFMs. By evaluating models prequentially on real future data in open-world environments, LiveHouse-TS shifts time series benchmarking from snapshot accuracy to continuous temporal validity. Rather than acting as a one-off leaderboard, our infrastructure serves as a continuous time series infrastructure designed to explore vital, long-term scientific questions: Can model rankings be maintained over the long term? Which models remain genuinely robust under distribution shifts? Extensive streaming evaluations across 11 domains with 17 datasets demonstrate that static rankings undergo a dramatic reshuffling under a live protocol.

Existing research on irregular time-series forecasting has primarily focused on model design, while evaluation metrics remain insufficiently studied. Existing benchmarks typically use mean squared error (MSE) as the evaluation metric. We show that, in irregular forecasting, MSE is determined not only by the model prediction but also by the sample-specific timestamp sampling distributions, leading to a biased assessment of the models' continuous-time predictive performance. To address this issue, we propose the Continuous-time Squared Error (CSE), which employs importance weighting to eliminate the influence of the timestamp sampling distributions. We further theoretically prove that CSE's asymptotic estimation error with respect to continuous-time risk is no greater than that of MSE. Finally, we construct a systematic benchmark covering synthetic, semi-synthetic, and eight real-world datasets to validate the effectiveness of CSE and systematically evaluate models' continuous-time predictive performance. Experiments show that CSE can recover continuous-time risk more accurately than MSE, while relying solely on MSE may not fully reflect models' continuous-time predictive performance in real-world scenarios. Our code can be obtained at https://github.com/hnu-vis/ITS-Bench.

Accurate 3D tooth segmentation forms the cornerstone of digital dentistry, yet it remains a formidable challenge due to the inherent intricacy of real-world dentitions, such as crowding, misaligned teeth and high morphological similarity between adjacent teeth. To resolve this, we present B-Spline Embedded Structure Learning, a novel framework that distills the inherent sequential arrangement of teeth into a continuous structural constraint to regularize representation space. Our approach parameterizes the global dental topology by fitting a parametric B-spline trajectory to tooth centers, assigning each point a continuous structural embedding that forces the shared backbone to capture global arch organization. To fully exploit these embedded priors, we introduce a Structure-Aware Dynamic Classifier (SADC) to substitute rigid static templates with adaptive, case-calibrated decision boundaries. SADC regularizes dynamic prototype pooling via a localized Gaussian proximity gate and contextually co-evolves them through an attention block modeling spatial relations and bilateral symmetries across teeth. Extensive evaluations on the 3DTeethSeg22 benchmark demonstrate that our method establishes a new state-of-the-art accuracy with exceptional structural robustness and efficiency in computational overhead, markedly enhancing the model's capacity to handle complex dental configurations.

GPT attention measures token compatibility through dot-product similarity. This mechanism is simple, effective, and memory-efficient. But it does not explicitly model whether strong token features should reinforce or suppress one another. We introduce Q-Interference, a fully classical quantum-inspired attention mechanism for autoregressive language modeling that augments each query and key feature with an amplitude and a learned phase. The resulting attention score is phase-aware which aligned phases contribute constructively while conflicting phases contribute destructively. Although Q-Interference yields a richer interaction rule than similarity alone, a naive implementation of Q-Interference requires a large token-pair-feature interaction tensor, making it memory-intensive and often impractical. To address this limitation, we propose an exact trigonometric factorization that computes the same score using two standard matrix multiplications avoiding materialization of the large intermediate tensor. Q-Interference fits directly into a Transformer block in GPT and leaves the remainder of the model architecture and next-token prediction objective unchanged. Experiments on public benchmark datasets and baseline models show that the proposed reformulation trains stably in a controlled GPT-style setting and provides a consistent memory advantage over naive phase-aware interference attention. These results support the specific contribution of this work: an exact memory-efficient reformulation that makes phase-aware interference attention practical within a standard GPT pipeline.

Curriculum learning has been widely adopted in the post-training of large language models by organizing training data from easy to hard. However, its effectiveness varies substantially across reasoning tasks, suggesting that no single curriculum is universally optimal and raising a fundamental question: what determines when curriculum learning works? In this paper, we answer this question by analyzing the optimization dynamics induced by different curriculum schedules. We show that the transfer relationship between different difficulty levels characterizes the optimization dynamics induced by curriculum learning, which in turn explains the effectiveness of different curriculum schedules, and formalize this relationship as Relative Transfer, a principled measure of cross-difficulty knowledge transfer. Based on this measurement, we derive Transfer-aware Dynamic Curriculum Sampling (TDCS), which dynamically adjusts the sampling distribution according to the estimated transfer relationship throughout training. Extensive experiments on multiple reasoning benchmarks demonstrate that TDCS consistently outperforms representative scheduling strategies across different tasks, model scales, and training paradigms. More importantly, our work provides a unified optimization-based explanation of curriculum learning through cross-difficulty transfer.

This paper considers robust nonadaptive regulation for general nonlinear systems in an output-feedback setting with arbitrarily high relative degree. We develop a nonadaptive design that combines an input-driven filter and a generic internal model with a recursive backstepping law, thereby recasting the regulation problem as the robust input-to-state stabilization of an augmented error system. Unlike adaptive schemes, the proposed method does not rely on linearly parameterized regressors and does not require the construction of Lyapunov functions having merely nonpositive derivatives. Under standard assumptions on the exosystem, including purely imaginary and simple eigenvalues, together with a minimum-phase input-to-state stability condition on the internal dynamics, we establish global asymptotic regulation and derive explicit, verifiable inequalities for selecting the design gains. The resulting nonadaptive framework guarantees convergence of the estimation and tracking errors even when the controlled-system dynamics are complex or only partially known. The effectiveness of the theoretical results is demonstrated using a benchmark controlled Duffing system.

Reinforcement learning (RL) has emerged as a powerful approach for improving reasoning in language and vision-language models, yet its strongest successes still depend heavily on ground-truth supervision (e.g., verifiable reward). Such annotations are costly to obtain and become increasingly scarce as reasoning capabilities advance beyond what humans can reliably evaluate. Self-rewarding RL reduces this dependence by enabling models to derive reward signals from their own completions. However, training solely on self-generated feedback can reinforce existing biases and suboptimal behaviors, reduce response diversity, and ultimately lead to homogenized responses and training collapse. In this work, we show that unsupervised reasoning can emerge through cooperative multi-agent training. We introduce Co-RL, a framework in which multiple decoupled models, sharing no parameters, are simultaneously optimized through RL using rewards derived from their peers. We further show that increasing cohort diversity, through heterogeneous model families, sizes, and rephrased training samples, reduces the correlated errors that drive self-reinforcing feedback loops. This diversity consistently improves reasoning performance, maintains behavioral diversity, and mitigates training collapse. Across text-only and multimodal domains, Co-RL consistently outperforms the base models and prior label-free approaches, while matching or surpassing supervised methods, without access to any ground-truth labels. Concretely, Co-RL yields average gains of 3.0-8.6% across seven text-only benchmarks for LLMs and 2.3-7.2% across four multimodal benchmarks for VLMs. Code is available at https://github.com/DrStranded/Co-RL.

Converting structural framing plans into editable finite-element model drafts remains labor-intensive and prone to transcription error. Existing drawing-understanding systems for building components rely on task-specific trained neural detectors, and language-model agents in structural engineering operate on text or model data rather than the drawing itself. This paper presents, to the authors' knowledge, the first framework applying an agentic vision-language layer to structural component detection and model drafting from framing-plan PDFs, without task-specific detector training or fine-tuning. A deterministic stage extracts primitives, estimates scale by dimension-ratio consensus, recognizes five entity classes with a drafting grammar, and assembles an editable layout. The agentic stage proposes typed corrections constrained by deterministic candidates, operation-specific admission tests, change-level review, and fail-closed transactions. Evaluation used an author-generated benchmark of 100 plans: a development half that informed every rule revision, and a seed-disjoint held-out half generated after the rules froze, evaluated once. All reported scores are end-to-end results of the complete framework on the held-out half. Scale was estimated within 0.1% of the generator reference for every drawing. Recall and precision were 0.922/0.997 for columns, 0.886/0.990 for beams, 1.000/1.000 for walls, 1.000/1.000 for braces, and 1.000/0.964 for openings. A controlled study repeated two corruptions three times on three development drawings. Calibration passed all nine trials; member repair met every strict end-state predicate in five of nine. Guarded review corrected missed framing and false marks within explicit bounds. The held-out half shares the development generator, so the study excludes independently drafted plans, raster evaluation, analytical connectivity, and solver validation.

HVAC systems represent a major share of building energy consumption. Traditional control strategies are limited in coordinating energy-comfort tradeoffs across multiple zones simultaneously. Reinforcement learning (RL) offers adaptive, data-driven control that optimizes performance over time. However, deploying learned neural network controllers in safety-critical building systems remains challenging due to lack of formal safety guarantees. We propose a safety-certified deep RL framework for multi-zone residential HVAC control. Proximal Policy Optimization (PPO) and Soft Actor-Critic (SAC) agents are trained in an EnergyPlus/Sinergym simulation to minimize energy consumption while maintaining thermal comfort. Post-training safety certification is performed on the PPO policy using Lipschitz-based forward invariance analysis, building on existing tools for the computation of Lipschitz constants for neural networks, to guarantee constraint satisfaction. Both agents are evaluated over an annual simulation cycle in an eight-zone variable refrigerant flow (VRF) testbed. The PPO agent achieves 67\% comfort violation reduction compared to rule-based control, while the SAC agent achieves 27.6\% energy savings. The PPO policy satisfies formal safety certification with a margin of $2.003^\circ$C. These results demonstrate the feasibility of combining reinforcement learning with post-training safety verification for multi-zone building control.

Financial-news direction prediction has become a popular NLP benchmark, yet reported gains depend critically on whether the train-test split is chronological or random, i.e., on temporal leakage. We audit this dependence on a 49,799-article corpus across 16 feature-model combinations spanning TF-IDF, MiniLM, FinBERT, and fine-tuned RoBERTa-large / DeBERTa-v3-large, plus separate zero/few-shot and LoRA probes of Llama-3 and Qwen2.5 LLMs: random splits inflate MCC by $1.1\times$ to $6.5\times$, tracking model capacity and feature richness, and end-to-end FinBERT fine-tuning re-amplifies rather than closes the gap (size-matched ratio $1.75\times$). Conditioning on event type, mergers and acquisitions (M&A) is the only audited category with a positive locked-test signal under near-temporal chronological evaluation (TF-IDF MCC $= 0.138$ train-only, $0.068$ under train$\cup$val refit; 10,000-permutation $p < 10^{-3}$); the signal does not transfer to FNSPID's 2009-2020 U.S. corpus, localising the headline to our 2024-2025 European-tilted M&A semantics rather than a universal predictor. Three independent role labellers converge on acquirer-tagged articles as the signal locus, a power-limited qualitative convergence rather than a hypothesis-tested asymmetry. Chronological splitting plays for financial NLP the role characteristics-purging plays for asset pricing: it strips the predictable, stale component of news and leaves a residual that is small, event-localized, and lexically shallow. We advocate leakage audits as a required disclosure for financial-NLP benchmarks.

End-to-end vision-language-action (VLA) and world-action models offer an elegant route to general-purpose robotics, but their reliability is bounded by validated physical coverage. When an unfamiliar object, sensor, embodiment, or contact falls outside that coverage and no validated fallback exists, correcting the failure requires new robot data, a policy update, and regression testing. This recurring burden is the retraining tax. Unlike text, embodied data must often be created by operating machines. We present Teach-and-Grow Learning (TGL), an agent-centered architecture for general robot learning. In its general form, a multimodal agent turns a few successful demonstrations into reusable Skill Blocks: closed-loop behaviors for meaningful subgoals. In a new scene, the agent grounds and composes these blocks, selects learned or geometric tools, observes the physical outcome, and revises the route when execution departs from intent. A Skill Library stores executable behavior, while structured Experience Memory carries forward success, failure, and repair. New tasks are acquired without task-specific policy retraining. Our LIBERO evaluation attains state-of-the-art performance; controlled studies expose skill induction, persistent reuse, and agent-directed adaptation. Finally, we propose the Teach-and-Grow scaling-law hypothesis: if X denotes effective reusable experience, future-task error and teaching demand should approach irreducible floors as power laws in X. The architecture therefore treats deployment as a period of continued learning, in which one task can make the next easier.

Vision-language models (VLMs) combine images and text, but when the two conflict and one becomes harder to read, it is unclear how a model shifts its reliance between them. We study this modality reallocation with a controlled setup: we degrade either the image or the text across four levels of legibility while keeping the other clean, and track how the model's preference changes. We build conflicts from GSM8K and SVAMP by pairing the rendered image of one arithmetic problem with the text of another, so the two sources support different answers. We also introduce ChartQA-Conflict, a manually reviewed benchmark of 229 chart-report conflicts with matched chart and table-image representations. We evaluate six open-weight VLMs using both generated answers and a length-normalized conditional log-likelihood margin. On GSM8K and SVAMP, five of six models shift more strongly away from degraded text than from degraded images. On ChartQA-Conflict, all six likelihood-scored models exhibit the opposite pattern, shifting more strongly away from the degraded visual source. This reversal persists after calibrating for unimodal accuracy loss and after replacing charts with plain table images. Two frontier API models, GPT-5.6-Luna and Gemini-3.5-Flash, behaviorally replicate the ChartQA-Conflict reversal, with GPT-5.6-Luna also matching the arithmetic direction. These results show that modality reliance in VLMs is not fixed, but varies across tasks, evidence structures, models, and evaluation settings. The source code is available at https://github.com/Ro-netizen004/multimodal-arbitration-artifact.

Multi-agent AI workflows are limited not only by model quality but by token cost, latency, and context-window quality. This paper presents a practitioner framework for token optimization and context-window management, grounded in an internal production dashboard that extracts structured work items from meetings, email, and chat with LLMs and routes summaries across workstreams. Six patterns are described: context stratification, fetch-once/process-locally architecture, schema-contracted prompts, token-aware fallback chains, semantic caching, and inter-agent communication compression. In production they cut measured cold-load latency to 61-116 seconds (six timed runs) from an operational baseline of roughly 3.5-10.5 minutes, with an estimated 60-70% token reduction. It also reports a controlled context-composition study: 2,420 confirmatory trials across 11 model configurations, using 661 anonymized workplace items scored for relevance. Holding the prompt at a fixed ten items, replacing some high-relevance items with same-domain low-relevance items improves the model's relevance-score concordance on the target items, versus high-relevance items only; we call this relevance-contrast context. In the all-11 paired analysis, the 50:50 signal/noise condition improved relevance accuracy by +0.077 over the 100% condition (naive 95% CI [+0.056, +0.098], Cohen's d = 0.49, Holm-adjusted p < .001, n = 220). These cells are not independent; by the nine model families the effect is +0.084 (95% interval [+0.064, +0.103]), reported as a within-corpus descriptive comparison, not a population inference. A Fusion-of-N follow-up found that learned synthesis did not beat the mechanical set union of item IDs. The contribution is a measured engineering layer between model research and production agent practice: repeatable patterns and evaluation methods for faster, cheaper, more reliable workflows.

Job Safety Analysis (JSA) and pre-task planning can benefit from prior incident records, yet historical accident data is often stored as unstructured narratives that are difficult to consult at the point of planning. A novel framework centered on large language models (LLMs) for highway construction safety reporting and planning is proposed as a foundation for future agentic applications, prioritizing deterministic, local inferencing. The first aim is to enable classification and quality scoring of incident narratives for existing and future reporting purposes. The second is to evaluate retrieval of relevant historical accidents, related imagery, and trusted industry documents for incorporation into daily safety plans. Neural probes were trained to classify incidents along four multiclass and two binary Occupational Injury and Illness Classification System (OIICS) fields and to derive an overall quality score, evaluated on a test set of over 15,000 narratives and a held-out set of 100 author-labeled records, benchmarked against a majority-vote LLM ensemble. The retrieval of historical accidents, reference imagery, and industry documents was benchmarked across embedding models using standard information retrieval metrics. OIICS classification reached 75% held-out accuracy, though the two binary flags were degenerate. The quality score, while meaningful on one database, was distorted on out-of-distribution fatalities in the held-out dataset. Accident retrieval recovered relevant incidents far above chance, performing best on lexically distinct construction activities. On document question answering, an open-weight decoder embedding model surpassed proprietary models. Overall, this work provides a new framework rooted in local inferencing and text embedding models for future agentic applications, with emphasis on bridging external data to JSA reports.

Video self-supervised learning through masked spatiotemporal prediction has emerged as a promising paradigm for learning feature representations from unlabeled data. However, existing methods typically rely on random masking, which indiscriminately removes regions irrespective of their semantic or temporal relevance. In ego-centric driving videos, this can weaken the pretext signal since safety-critical cues such as pedestrians, vehicles, lane boundaries, and dynamic interactions often occupy only a small portion of the frame, yet are central to downstream perception. We introduce V-JEPA4A, a domain-specialized variant of V-JEPA for autonomous driving that is pre-trained on publicly available driving videos with a novel saliency-driven masking policy. It accounts for semantically and temporally relevant context. The proposed policy preserves and predicts context according to semantic importance and temporal relevance, yielding more informative representation learning while retaining the efficiency of masked prediction. We evaluate the resulting encoders on four driving benchmarks spanning tracking, semantic segmentation, and depth estimation. The results demonstrate that V-JEPA4A reduces identity switches on BDD100k MOT by 25% over V-JEPA with random masking, achieves 73.2 mIoU on Cityscapes, and 3.75 RMSE on KITTI-2015 depth, while incurring only ~14% additional pre-training iteration overhead.

Online decision making often requires navigating a landscape shaped by both dynamic contexts and strategic interactions. In competitive pricing, for example, hotels must account for both dynamic contextual factors and rivals' strategic responses. Existing approaches address only part of this challenge: contextual bandits optimize single-agent decisions using observable features but ignore multi-player interactions, while online matrix games capture strategic behavior through Nash equilibrium but assume fixed payoffs, ignoring contextual information. How should agents act then when strategic payoffs evolve with contextual signals? We introduce \emph{online contextual matrix games} to integrate contextual information into multi-player online games. We further propose \emph{OnGameLearn}, an online learning algorithm that efficiently balances exploration and exploitation across both player actions and contexts. This approach comes with statistical guarantees: tail bounds for the estimated payoff matrix, the convergence of the estimated Nash equilibrium, the asymptotic normality of the parameter estimators, and the sublinear regret bound. We also develop the notion of \emph{policy value} in matrix games and develop a doubly robust, $\sqrt{T}$-consistent estimator for it. Across simulated studies and a real-world hotel pricing application, we find that OnGameLearn effectively navigates the intertwined challenges of strategic and contextual decision-making.

Many table-centric NLP tasks such as NL2SQL first retrieve relevant tables from large collections using keyword search. Recent work uses LLMs to generate natural-language table descriptions to improve retrieval, but they are typically optimized for fluency rather than retrieval effectiveness. We present Polaris, a system that trains an LLM to generate table descriptions directly from retrieval feedback. Our key insight is that existing table retrieval benchmarks already contain the supervision needed for this task: given query-table relevance judgments, we generate multiple candidate descriptions for each table, rank them by their BM25 retrieval effectiveness, and use the resulting preference pairs to fine-tune the LLM with Direct Preference Optimization (DPO). Polaris further expands abbreviated table and column names before generation to reduce vocabulary mismatch. Extensive experiments show that Polaris outperforms the state-of-the-art AutoDDG solution, often by a significant margin. More broadly, our results demonstrate that retrieval benchmarks can be repurposed as supervision for training LLMs to generate retrieval-oriented metadata.

Reasoning has become a standard technique and feature for contemporary LLMs; however, its application and quality in the context of demanding legal-oriented tasks, such as legal case forecasting, remain under explored. We investigate how LLMs reason in the context of legal case forecasting, using legal cases from the European Court of Human Rights (ECtHR) as a testbed. We evaluate OpenAI GPT 5.4, a recent top-tier LLM, by exploring alternative prompting strategies that are more or less suggestive of what counts as legally meaningful reasoning in the context of ECtHR jurisprudence. We present our findings derived from assessing the model's responses with both human and LLM evaluation. We find that the examined model scores far from ideal in legal reasoning, the model produces structurally complete but substantively shallow analyses, and that LLM-as-a-Judge evaluators are internally consistent yet align only weakly with our trained annotators, i.e., reliable but not a valid substitute for human evaluation. Overall, the expert-curated prompt leads to more comprehensive reasoning, which does not result in more accurate predictions compared to the other examined settings. Based on our findings, we urge the community not to rely solely on automated LLM-based evaluation and to avoid using task accuracy as an appropriate proxy for reasoning quality.

Active inference selects actions by minimising an expected free energy functional over predicted futures. However, adding an expectation over yet-unobserved outcomes means the free energy functional no longer has a Kullback-Leibler structure, which hinders message passing treatments of inference procedures. We propose an alternative formulation based on a Bethe free energy functional, fully supporting inference by message passing. The epistemic drive is maintained by imposing an information constraint, next to normalisation, marginalisation and form constraints, insisting that the mutual information between future observations, states and parameters given actions must be at least as large as the entropy of the goal prior. For a specific value of the corresponding Karush-Kuhn-Tucker multiplier, the stationary point of this constrained Bethe Lagrangian recovers the expected free energy solution. We show that, as the information demand is varied, the solved multiplier moves through its inactive, interior, and saturated regimes. In the inactive regime the agent's epistemic drive switches off entirely, while in the saturated regime it is maximal. We compare the performance of the constrained Bethe agent on three tasks against EFE and Q-MDP.

Textual context such as news, reports, and logs can provide valuable signals for time series forecasting, especially when future dynamics are driven by external events that are not yet visible in historical values. Existing multimodal forecasting methods often either ask large language models (LLMs) to predict numerical values directly or fuse text and time series implicitly, making contextual influence difficult to interpret and control. We propose SCENARIODIFF, a hierarchical contextual reasoning framework for multimodal time series forecasting under noisy and weakly aligned documents. SCENARIODIFF organizes contextual information into three levels: a Historical Context Agent extracts stepwise evidence from raw documents, a Scenario Agent produces a qualitative scenario description for the forecast horizon, and an Anchor Guidance Agent generates sparse anchor points for event-relevant future regions. These structured signals condition a Multimodal Diffusion Transformer, while Anchor Blended Sampling locally refines generated trajectories without retraining. Experiments on the Time-MMD benchmark show that SCENARIODIFF is especially effective in event-driven domains, demonstrating the value of explicit hierarchical scenario guidance for multimodal time series forecasting. Our full implementation is available at https://anonymous.4open.science/r/ScenarioDiff_ICDM-2C4C

Off-policy reinforcement learning (RL) has become increasingly sample-efficient, enabling applications such as RL fine-tuning of Vision-Language-Action models into reliable, high-performing policies. World models offer a further lever for sample efficiency, as they predict state changes rather than actions alone, but their success has largely been confined to supervised policy learning. Prior model-based RL methods often optimize the policy or value function directly on imagined rollouts, which is prone to compounding bias and struggles to scale to large, high-dimensional problems such as real-world robotics, a problem that worsens with task horizon and visual complexity. In this work, we instead ask whether we can leverage world models directly on top of standard Q-learning to improve performance, while remaining trained and grounded in the real, online setting. We propose QWM, a framework that leverages world models to perform test-time search over imagined trajectories on top of Q-learning to select high-value actions during both online rollouts and evaluation. Since the policy and value function are trained only on real transitions, QWM avoids compounding model bias while still gaining the sample-efficiency benefits of predictive search. On challenging manipulation benchmarks Robomimic and LIBERO, QWM significantly outperforms strong prior state-of-the-art methods on both sample efficiency and performance.

What a language model internalizes from fine-tuning is usually diagnosed after the fact. We make it an experimental variable. OraclePhys is a systematic fine-tuning framework with three components: OraclePhys-Bench, an exactly-graded structural-mechanics benchmark whose finite-element oracle scores every answer and counterfactual edit -- no human labels, no LLM judging; OraclePhys-30K, a supervision dataset of seven answer forms over byte-identical structure descriptions; and a controlled training study across the seven forms and three verifier roles. The study yields two findings. First, the label's answer form -- not its bit count -- causally determines what fine-tuning teaches: a ranking objective installs an out-of-distribution forward model where the untrained base sits at the guessing prior, a scalar objective at best a partial one, a boolean nothing detectable; the vector-scalar gulf survives a second physics domain, a second model family, and a paraphrased evaluation surface. Second, written or score-filtered answers install this capability, while advantage-weighted scores (GRPO) raise reward yet leave the model statistically equivalent to its start on held-out physics -- within the recipes and budgets tested -- sufficing only for routing. The trained 8B -- the first LLM on spatial structural response -- reaches the task's data-precision frontier: above a frontier LLM at zero- and 32-shot, at a specialist's level. What the label spells out about the target computation is what fine-tuning teaches; what you train on is what you route.

Retrieval-Augmented Generation (RAG) has significantly enhanced the performance of large language models (LLMs), yet these systems remain vulnerable to knowledge-poisoning attacks, in which misinformation in retrieved documents can influence the model's final outputs. Notably, an LLM may correctly detect that a document contains incorrect information while nevertheless being influenced by it. Prior work has addressed this vulnerability through the Cordon Principle, which prevents models responsible for final answer synthesis from directly accessing raw evidence. Although effective, this strict isolation can introduce substantial computational overhead. In this work, we propose a refined security principle: only agents capable of deliberative System 2 reasoning may access untrusted documents. To evaluate this principle, we introduce novel metrics that quantify the discrepancy between misinformation detection and downstream influence. We then empirically compare state-of-the-art reasoning language models with standard language models across these metrics. Our results show that reasoning-capable models are substantially more robust to corrupted evidence, without requiring the strict isolation imposed by the Cordon Principle. These findings provide empirical support for our refined principle and suggest a more practical foundation for secure RAG system design.

To effectively collaborate with users on knowledge-intensive tasks, Large Language Models (LLMs) must perform information calibration: matching content to a user's evolving understanding and cognitive capacity. Yet user simulators used to evaluate and train LLMs do not explicitly model user knowledge so they neither produce realistic interactions across knowledge levels nor reflect how interactions unfold as that knowledge evolves. To close this gap, we introduce KNOWSIM, an evaluation framework built around a user simulator that maintains explicit knowledge states, represented as a graph of Information Units with prerequisite relationships, that evolve under update rules grounded in learning theory. KNOWSIM computes three metrics (Knowledge Gain, Delivery Calibration, Cognitive Overload) directly from the knowledge state trajectory, reflecting key mechanistic aspects of information calibration. We validate KNOWSIM against 705 human-AI sessions across two domains, stratified by knowledge level: its rankings align significantly with human judgments (73-74% sign agreement), outperforming three baseline simulators. Applied to 9 LLMs, KNOWSIM reveals that the best model shifts by user knowledge level, revealing aptitude-treatment interactions invisible to standard evaluation.

Symbolic planning with PDDL offers a principled framework for long-horizon robot manipulation, but constructing accurate PDDL domain and problem descriptions remains a significant bottleneck, typically requiring substantial domain expertise. We present a Vision-Language Model (VLM)-based approach called PDDL-ART, a framework that autonomously generates task-specific PDDL domain and problem descriptions from a single expert demonstration, a natural language task description, and a library of available high-level action names. PDDL-ART does not require any domain templates, action signatures, or fine-tuning. To ensure the generated descriptions are not only syntactically valid but semantically aligned with the demonstrated task, PDDL-ART introduces a multi-stage correction pipeline operating at syntactic, semantic, and execution levels. A key component of execution-guided correction is symbolic predicate grounding. Instead of relying solely on visual observations, PDDL-ART leverages the tool-use capabilities of modern VLMs to incorporate geometric and temporal reasoning for evaluating relational predicates that are not directly discernible from images alone. Critically, the model autonomously determines when to invoke these tools and how to interpret their outputs. We evaluate PDDL-ART on challenging manipulation tasks in engine maintenance and household domains, including tasks that require memory, abstract predicate inference, and goal states that are visually indistinguishable from the initial state. PDDL-ART achieves an average success rate of 93.3%, compared to 78.3% for a baseline VLM-based planner.

Vision-language Models (VLMs) excel at 2D grounding, spatial reasoning and agentic tool-based planning in static scenes. However, consider asking a home robot "Is my medication still in the cabinet?" The answer may be physically hidden behind a row of containers that must first be moved aside. Answering such questions in real-world cluttered environments requires reasoning in dynamic scenes: distractors must be manipulated to reveal occluded objects, and each action changes the scene the model must reason over. We formalize this setting as Manipulation-Grounded Visual Question Answering (MG-VQA) and introduce PROBE, a framework for benchmarking and finetuning VLM agents on such tasks. We first develop PROBE-Sim, a high-fidelity tabletop simulator with everyday objects and a robot manipulator equipped with grasping and pushing tools. PROBE-Sim is used to create PROBE-Bench: an evaluation suite of 150 tasks across 6 question types on cluttered tabletop scenes, where a VLM perceives, picks up or pushes objects before answering. We observe consistent trend across all frontier VLMs: agentic tool-based methods outperform their perception-only baselines (8.0% on average) across all task types. We further design PROBE-Agent, a finetuning recipe to distill successful trajectories from a powerful teacher foundation model to a smaller open-weight model using a mixed data recipe that encourages manipulation-efficient question answering. PROBE Agent finetuned models outperform their off-the-shelf agent baseline (11.5% on average) and demonstrate positive transfer to unseen objects and a held-out task. We validate sim-to-real transfer by deploying PROBE-Agent finetuned policies in real-world tabletop environments.

Open-vocabulary monocular 3D detectors report strong in-domain performance, but each evaluates under a different protocol, several rely on per-image category oracles unavailable at deployment, and all collapse geometry and semantics into a single AP metric. To address this, we introduce OV3D-Bench, a diagnostic benchmark that compares open-vocabulary monocular 3D detectors under deployment-realistic conditions across seven indoor and outdoor datasets. Our benchmark replaces the per-image class name oracle with test-time dataset-level class name prompts, and decouples detection accuracy along three axes: localization, semantic robustness, and cross-domain transfer. We evaluate seven representative detectors and find that (i) they localize objects well yet often mislabel a correctly localized box as a semantically adjacent category; (ii) accuracy is highly sensitive to prompt phrasing (e.g. WildDet3D's performance collapses from 18.6 to 5.4 AP when prompted with "a detailed high-resolution photo of a car" rather than "car"); and (iii) the widely adopted target-aware protocol hides these errors (e.g. inflating DetAny3D's AP by 1.9 $\times$ on ScanNet). Lastly, we demonstrate that simply remapping a frozen closed-vocabulary detector's predictions using a contrastive vision-language encoder such as SigLIPv2 performs competitively against recent purpose-built open-vocabulary methods. This indicates that geometric localization is more mature, while open-vocabulary semantics remains the primary bottleneck.

Vision-language-action (VLA) driving models couple a reasoning stage with a diffusion-based trajectory decoder, but do not give a direct way to redirect attention toward safety-critical actors at inference time without retraining. We studied a bounded additive pre-softmax attention bias on the visual tokens of detector localized traffic actors on Alpamayo-R1's Qwen3-VL backbone. It is applied as a fail open forward pre-hook with no weight changes. On 50 lane-change scenarios from the Physical AI World Model Synthetic dataset. The trajectory decoder shows a monotonic dose response in the bias magnitude, separate from a paired zero bias control at every tested magnitude. It reaches $\approx 17$\,cm mean displacement with lateral shifts up to $\sim 140$\ cm at the clamp. A layer ablation places the action-relevant signal in late layers, where the effect increases with the number of hooked layers (2.0cm for the first 8 layers; 67.6cm for all 36). A per call injection audit explains why the Chain-of-Causation text never changes. The mask based bias never reaches the reasoning pathway in this serving stack, so the invariance is verified exposure, not robustness. Steered trajectories tend to shift toward the attended actor, suggesting the bias governs where the model looks rather than encoding a target behavior.

A central goal of multilingual NLP is to achieve high monolingual performance per language and cross-lingual alignment for large-scale language coverage with a multilingual model. The curse of multilinguality describes the phenomenon of degradation in multilingual model performance as we increase language coverage, posing a threat to the above goal. This paper asks whether multilingual embedding spaces are inherently incapable of achieving perfect multilinguality without a prohibitive increase in required capacity. We first formalize the goal of "perfect multilinguality", embodied in two multilinguality conditions. We then prove that the minimum dimensionality required for perfect multilinguality grows only logarithmically in the number of languages. That is, we show that there is no theoretical curse of multilinguality for embedding space structure. This suggests that the empirical curse of multilinguality is a result of real world data and training conditions. We back this understanding with a small-scale empirical study. Our paper provides the first theoretical and intrinsic perspective on the curse of multilinguality, with implications for the scientific understanding of this phenomenon.

Multimodal large language models (MLLMs) increasingly answer questions whose correctness depends on visual, textual, temporal, acoustic, document, chart, or embodied evidence. Their failures are therefore not only linguistic. A fluent answer may conceal poor input quality, a perceptual error, weak grounding, conflict between modalities, unstable reasoning, distribution shift, or a question that is not answerable from the supplied evidence. This survey organizes the literature on uncertainty-aware MLLMs around a decision-centered framework: uncertainty sources give rise to observable signals, signals must be calibrated or controlled for risk, and calibrated uncertainty should determine the system action. We review work on token and logit uncertainty, semantic disagreement, perturbation instability, grounding and attribution scores, verbalized confidence, verifier and judge scores, conformal prediction, selective answering, abstention, clarification, retrieval, self-checking, and escalation. The central argument is that uncertainty should not be evaluated only as a confidence number; it should be evaluated by whether it improves behavior under insufficient, conflicting, shifted, or high-risk multimodal evidence. We position this survey against text-only uncertainty and abstention surveys, broad MLLM surveys, MLLM hallucination surveys, and safety-oriented reviews. We conclude with open problems in source-aware decomposition, action-aware benchmarks, calibration under shift, black-box uncertainty estimation, broader modality coverage, reproducible reporting, and human-centered uncertainty communication.

Next-encounter ICD forecasting predicts which standardized diagnosis codes will be documented at a future visit from the longitudinal record available beforehand. The task is prospective and multi-label: the target note does not yet exist, and several codes may be correct. Structured EHR foundation models capture recurrence and temporal progression, whereas language foundation models generate flexible diagnostic hypotheses. We introduce ICD-Deepresearch, a DeepResearch workflow that composes these predictive foundation models with medical search and ICD dictionaries. Because no source reveals the future code set, research evaluates candidate transitions by linking patient evidence, external clinical relations, and exact code semantics under a fixed top-K budget. Candidate Generation uses SparseEHR to produce an EHR Prior that initializes two bounded Research Expansion rounds; an independent GPT-5 Direct Forecast supplies complementary candidates. Final Selection validates, deduplicates, and jointly ranks both paths, after which a separate module writes rationales without changing predictions. Finally ICD-Deepresearch achieves patient-averaged precision/recall of 24.60/35.09% on MIMIC-III and 25.14/48.32% on MIMIC-IV. Physicians rate 51% and 68% of its retrieved documents useful, compared with 22% and 39% for standalone GPT-5 web search and 32% and 41% for Medical Deep Research. ICD-Deepresearch therefore improves over the registered local comparators while retrieving evidence with higher physician-rated usefulness than the standalone research systems

We present KernelArc, a multi-agent framework for autonomous GPU kernel optimization across heterogeneous workloads. Strategy-specialized agents run in parallel and coordinate through conclusions-only shared memory, a deterministic benchmark guard, and read-only cross-agent state with plateau-triggered drafting. We evaluate \kernelarc{} on NVIDIA H100 and B200 GPUs using category-representative SOL-ExecBench workloads. The resulting implementations span custom BF16 GEMM, static cuBLASLt Expert-API configuration tables, fused mixture-of-experts backward, shape-gated decoder-layer fusion, native NVFP4 grouped-query attention, and paged prefill attention. At the public SOL-ExecBench leaderboard snapshot recorded on July~30, 2026, these submissions ranked first on representative L1, L2, Quantization, and FlashInfer tasks. The trajectories support the paper's central motivation: shared multi-agent search can broaden exploration and reach stronger incumbents within a fixed candidate budget, while the value of individual coordination features depends on the kernel and optimization stage.

Distinguishing a genuine foul from a simulated dive in football remains one of the sport's most contested fine-grained recognition problems, especially when such decisions have to be from a single broadcast view without multi-view camera angle. We introduce a balanced 600-clip single-view Foul/Dive dataset and show that contact-aware sampling concentrating the model's attention around the moment of physical contact rather than treating all frames equally yields substantially improved recognition of this contact- specific problem. The proposed approach achieves 86.0% accuracy and macro-F1 0.860 on the held-out test split, a 12 percentage- point gain over contact-unaware alternatives that grows further on unseen data. We also evaluate each pipeline component against human annotations, establishing where and why the system suc- ceeds and fails. The result is a documented dataset, a reproducible single-view pipeline, and a grounded evaluation framework for fine-grained contact-event recognition in broadcast football footage. The dataset and code are available at https://github.com/hossain- tamim/contact-aware-dive.

A bounded agent may obtain information for a decision from its own past, from peers, or from both sources. Retaining task-relevant history can reduce later communication, while a peer message can supply what memory lacks. Under limits on both resources, how should an agent allocate its information budget? Given a fixed task and decision rule, the memory and message rate pairs attaining a performance threshold form an achievable region under specified rules for using history and peer observations. We call its efficient boundary the remembering--signaling frontier. Across conditions where history permits the same maximum reduction in task loss, we hypothesize that a bounded agent will need less peer communication when it obtains a larger loss reduction from history. In preliminary referential games, target repetition coincided with shorter successful messages, while predictability from a hidden cyclic rule did not shorten them. Experiments varying memory and message rates can estimate the frontier and test this prediction across cooperative tasks.

Secondary use of electronic health records requires de-identification, yet existing systems miss \emph{institutionally situated} protected health information (PHI) such as hospital abbreviations, building names, and internal codes whose status is locally determined. We ask whether large language models (LLMs) with in-context learning (ICL) can close this gap and control the precision--recall trade-off. On 100 annotated pediatric oncology notes (5,322 PHI spans) from Texas Children's Hospital, we benchmarked eight LLMs against two purpose-built systems (Stanford TiDE, OpenMed PII) and two pattern-based baselines. Each LLM ran under three prompts of increasing specificity: (1) a HIPAA-aligned baseline, (2) baseline plus the institutional PHI categories it missed, and (3) prompt 2 plus instructions against over-redacting clinical content. We then compared 14~multi-agent and ensemble configurations against the best single prompt, with recall the primary safety metric. LLMs outperformed the purpose-built systems (best F1=0.918$\pm$0.001 vs.\ TiDE 0.779), with advantages concentrated in contextual categories. Naming the missed categories recovered 79\% (48/61) of them, and discouraging over-redaction restored precision. No agentic architecture beat calibrated single-pass prompting (F1 0.906--0.907), but LLM outputs surfaced 414~candidate annotation gaps; re-annotation confirmed 227~PHI spans, against which the final prompt reached recall=0.981 (F1=0.907$\pm$0.002). Well-calibrated ICL resolves both the institutional PHI gap and the precision--recall trade-off in one LLM call per note. LLMs cost more to run than traditional methods, but that cost buys a way to audit the reference standard. LLMs are a legitimate, adaptable alternative to purpose-built de-identification systems; institution-specific prompt development should be the primary adaptation strategy.

Methods for improving knowledge use in large language models typically fall into two regimes. Non-parametric retrieval offers flexible access to external knowledge, but adds retrieval latency, context overhead, and only shallow integration with the backbone. Parametric adaptation is efficient at inference time, but entangles knowledge with model weights and can be hard to update, audit, or transfer. Engram-style hashed memory occupies a middle regime: it stores learned information in an external, addressable table, yet consumes that table through a small learned reader. This raises a basic question: when such a memory is moved across backbones, what matters more, the frozen memory itself or the target-side reader? We study this question through cross-model frozen-memory extraction, in which a memory trained on a source model is frozen and attached to a different target model, with only a lightweight reader trained. Ablations show that learned memory content and correct addressing both matter, but the transferred table becomes useful only through a reader aligned to the target model. In downstream question answering tasks, a dual-layer, four-branch reader nearly closes the gap between same-model and cross-model reuse, achieving an average score of 38.8 under our controlled evaluation protocol. Moreover, when the provider reader is directly compatible with the target interface, the frozen artifact can provide substantial utility without target-side training, while optional reader adaptation yields further improvement. These results suggest that Engram can serve as a reusable external knowledge artifact, provided that the target has access to a compatible reader interface; target-side adaptation can further improve alignment when direct reader reuse is insufficient.

The 10th AI City Challenge, held with ECCV 2026, marks a decade of community benchmarking for intelligent transportation, smart cities, and physical AI. Since its 2017 start with vehicle detection, classification, and tracking, the challenge has grown into a broad benchmark suite for multi-camera perception, multimodal reasoning, synthetic-to-real learning, generative forecasting, and privacy-preserving evaluation. The 2026 edition continued this growth with 325 registered teams, up from 245 in 2025, and participation from 26 countries and regions, up from 15. Its six primary tracks cover multi-camera 3D perception, transportation safety captioning and VQA, traffic anomaly reasoning, text-based person anomaly search, generative traffic video forecasting, and cross-city object detection. Track 3 further includes two out-of-domain leaderboards, submitted as Tracks 7 and 8, for fisheye traffic-violation understanding and pedestrian situated-intent VQA. This paper summarizes the challenge setup, datasets, evaluation protocols, leaderboard results, and workshop papers. Across tracks, successful systems combine foundation models with geometric grounding, retrieval or reranking, synthetic-data design, domain adaptation, and controlled inference.

Visual loco-manipulation policies that can generalize to novel scenes and objects have long been a goal of robotics research. However, today's data-hungry algorithms make collecting sufficient demonstrations a struggle for tabletop manipulation, and even more so for humanoids that must also walk and balance. Learning from simulated data and transferring that behavior to the real world, as is commonly done in locomotion, sidesteps this struggle, so we replicate that recipe for loco-manipulation. In doing so, we find that cloning synthetic demonstrations results in a low performance ceiling no matter the amount of training data. Reinforcement learning breaks through it, and refining the cloned policy with Flow-GRPO on a single sparse reward yields performance that synthetic behavior cloning cannot match. Together, these stages form our end-to-end sim-to-real pipeline spanning more than 150,000 scenes, which we use to train FetchMan. We evaluate it on FetchMan-Bench, a simulation benchmark we release, and deploy it zero-shot on a real Unitree G1, where our single-object reach-and-pick policy walks to and grasps a target across unseen scenes at 73.3% success. Finally, we extend this recipe to multi-object training, a first step toward loco-manipulation generalist policies at this data scale.

Recent monocular depth estimators achieve strong zero-shot generalization, yet often struggle to preserve fine-grained structures and object boundaries. We attribute this limitation to the prevalent combination of large-patch ViT encoders and convolutional decoders, as coarse tokenization can weaken pixel-level cues that upsampling cannot fully recover. To address this issue, we propose PXDepth, a discriminative monocular depth model that separates global context modeling from pixel-level depth prediction. Specifically, a large-patch ViT captures global scene context, while a pixel-space predictor composed of Context-Modulated Pixel Transformer blocks maintains high-resolution spatial representations throughout depth estimation. This design preserves fine structures and sharp boundaries without sacrificing global depth consistency. Across diverse zero-shot benchmarks, PXDepth combines faithful local geometry with competitive global depth accuracy while remaining efficient at inference. Our code and model are available at https://yuanzhy29.github.io/PXDepth-Page/.

Long-horizon robot manipulation chains many contact-rich skills into one multi-stage task. Vision-language-action (VLA) models increasingly master the individual skills, yet the chain still fails: errors compound beyond the policy's ability to correct, and one subtask silently constrains the next. A promising recipe freezes the VLA and puts an LLM agent in charge: it plans in language, moves in free space with analytic primitives, invokes the VLA only for contact-rich segments, and writes adaptation into language memory. Applied to long horizons, it breaks twice. (1) Competence comes from whole-task exploration at test time, whose cost is multiplicative in stages: if one stage needs T episodes, a K-stage task needs about T^K, and a failure does not reveal which stage caused it. (2) It has no representation of transitions: the VLA primitive carries an exit but no entry condition, so a subtask can succeed in a form its successor cannot use. We present BATON. Against (1), BATON makes the subtask the unit of exploration: each is explored in the cheap short-horizon regime and its solution stored in memory; a long-horizon trajectory is then composed from these solutions rather than discovered whole. Cost becomes additive (T*K) and every failure is attributed to a single stage. Against (2), BATON equips exploration with a transition-aware memory. Within a subtask, a verifier agent governs the invocation transition: the VLA is called only after the wrist view confirms the scene is ready. Across subtasks, a handoff transition restores an entry state disturbed by the predecessor's residue, and a lookahead transition selects the strategy whose outcome the successor can inherit. No parameters are updated. On the long-horizon benchmark RoboMemArena, BATON improves task success by 11.6% and cumulative success by 14.9% over the SoTA.

This paper investigates an increasingly important topic in generative modeling: pixel-space diffusion models. Although numerous studies have explored this topic, most focus on small-scale or class-conditional settings. Consequently, a practical recipe for training pixel-space models that rival or exceed well-established latent-space counterparts remains elusive. Through a comprehensive empirical study, we first observe that direct large-scale pre-training in pixel space converges substantially more slowly than in latent space. This observation motivates a latent-to-pixel strategy that acquires generative priors efficiently in latent space and transitions to pixel space during post-training. We then systematically investigate the key design choices governing this transition, including weight initialization, data composition, prediction target, decoder architecture, and noise schedule, and identify a practical recipe that makes the resulting pixel-space models match or outperform their latent-space counterparts while delivering 3.18 to 4.75 times end-to-end inference speedups. We hope that our findings provide useful empirical insights and practical guidelines for future research on pixel-space generation.

Long-horizon robot manipulation requires a robot to both execute individual skills reliably and sequence them coherently over extended tasks. Most hierarchical vision-language-action (VLA) models make each such decision with a single forward pass, leaving no mechanism to allocate additional computation to difficult or consequential choices. We introduce $τ_0$-VLA, a hierarchical robot foundation model that formulates high-level subtask generation as a compute-scalable inference problem through world-model-guided test-time computation. At each inference step, the high-level policy uses execution memory to generate a subtask and, when needed, searches over alternatives before committing to its output. A low-level policy then executes the generated subtask across multiple robot embodiments. The policy is trained on 40,115 hours of heterogeneous real-world data with multimodal co-training. Across in-domain and distribution-shifted settings, allocating additional test-time computation substantially improves next-subtask prediction accuracy, and these gains translate into higher closed-loop success on long-horizon robot manipulation tasks.

The quadratic cost of attention-based sequence models for long contexts has motivated a growing line of research on memory-based models that can compress context into a compact state. However, most existing memory models expose a static memory throughout the entire sequence. Because early tokens face no compression pressure, they occupy too many degrees of freedom and "pollute" the memory state, leaving little capacity for later context and increasing interference between what is stored and what arrives next. We study a new paradigm of incremental memory activation, where the effective capacity of memory is progressively expanded as the context grows. Imposing an early bottleneck forces the model to compress history more effectively, while unlocking fresh capacity over time reduces interference and improves retention of later context. We instantiate this paradigm in Proteus, a straightforward mechanism that can be incorporated into a broad class of neural memory architectures at no additional cost. We apply Proteus to state-of-the-art models, including SWLA, Comba, Titans, and Hope-Attention, and observe consistent improvements on standard language modeling and reasoning, as well as on long-context retrieval and understanding, with gains that grow at longer context lengths. Overall, our results show that static memory is suboptimal and that scheduling effective capacity is a simple and broadly applicable tool for sequence modeling.

Foundation models are increasingly used for perception, reasoning, planning, and action generation in embodied agents, creating security risks that can propagate from digital inputs to physical behavior. Existing surveys often organize threats by mechanisms such as jailbreaks, prompt injection, backdoors, poisoning, or adversarial examples, but these categories do not consistently identify where an adversary first enters the embodied control loop. We present a trust-boundary-centric survey of foundation-model-powered embodied-agent security. Using a first-compromised-trust-boundary principle, we separate attack surface from attack mechanism and organize the system into five layers and twelve attack surfaces spanning the model supply chain, user instructions, context and memory, physical semantic environments, multimodal perception, world state, internal reasoning, task planning, action interfaces, middleware, multi-agent communication, and execution control. Based on 58 attack records and 61 defense records collected through August 15, 2026, we analyze representative attacks, cross-layer propagation, defense placement, and evaluation practices. Our quantitative analysis shows that attack research is concentrated on multimodal perception and action interfaces, while defenses are especially concentrated on action-level and runtime protection. Context and long-term memory, middleware and networking, world-state integrity, and multi-agent trust remain comparatively underexplored. We conclude with open challenges in state provenance, compositional defenses, long-horizon attack propagation, physical realizability, Byzantine multi-robot behavior, and unified closed-loop evaluation.

Turning a frontier vision-language model into a robot policy usually means fine-tuning it to emit an action representation it never saw in pretraining, which throws away much of the reasoning that made the model worth reaching for. We go the other way and keep the VLM frozen. It writes the policy as a short Python control function, with no demonstrations and no fine-tuning. Writing that code once is open-loop, though. Existing closed-loop methods react at the wrong level: they retry a fixed policy or pick a different subtask, but never rewrite the code that failed. VLCP closes the loop where the failure actually lives, on the control code, within a single episode. Every $K$ steps the VLM re-observes the scene from multi-view RGB, proprioceptive state, and a state delta, then rewrites the control function from what it just saw, so a failure is caught before it compounds. We evaluate on a 57-task MuJoCo/RoboVerse sweep. This training-free policy reaches $35.1\%$ pooled success, against $3.5\%$ for the identical system queried once per episode. That tenfold gap holds with non-overlapping confidence intervals in every scene family. The gain traces to a $27.3\%$ within-episode recovery rate on failed grasps: a miss an open-loop controller would carry to the end of the episode gets re-observed and fixed at the next replan. And the loop stays cheap. A median $84\%$ of input tokens hit cache, an episode needs only about $10$ compact queries, and control blocks written during any replan persist to a cross-episode skill library reused in later prompts.

Humanoid robots hold great promise as general-purpose agents in human-centered environments, yet generalist vision-language-action (VLA) foundation models are not readily applicable to humanoid whole-body loco-manipulation. The high dimensionality and interdependence of humanoid motions make it challenging for conventional single-stage VLA architectures to coordinate locomotion, waist posture, and dual-arm manipulation effectively. Moreover, policies trained through offline behavior cloning can remain suboptimal during real-world deployment. Although online reinforcement learning can refine policies through real-world interaction, directly tuning large VLA backbones demands excessive computation and may introduce safety risks during real-robot exploration. To address these bottlenecks, we introduce HAF (Humanoid Adaptation Framework), a two-part framework consisting of HAF-VLA and HAF-Steer that transfers off-the-shelf generalist VLA foundation models to humanoid whole-body loco-manipulation. HAF-VLA is a hierarchical action-flow generator built on a pretrained flow-matching VLA. It splits full-body action denoising into three sequential stages with stage embeddings and cross-stage KV caches that retain kinematic dependencies, avoiding incoherent whole-body actions from one-shot generation. On top of the frozen HAF-VLA, HAF-Steer is a latent offline-to-online RL pipeline that leverages flow-matching invertibility and DCT-based dimensionality reduction to restrict RL optimization to a compact noise subspace and train a regularized SAC policy. This avoids updating the large VLA backbone and enables efficient real-world policy refinement. Evaluated on seven real-world humanoid loco-manipulation tasks, HAF surpasses vanilla single-stage VLA baselines and improves whole-body coordination and task performance. Project website: https://grange007.github.io/HAF .

Generative pretraining established reusable task representations; later work on language-based task conditioning and in-context learning showed that a fixed model could adapt its behavior from instructions and demonstrations. Policy Iteration with Human Feedback (PIHF) builds on this development and the recurrent evaluate-and-improve structure of generalized policy iteration. PIHF uses a pretrained language model as its execution substrate and moves persistent revision to a versioned natural-language policy and tool set. A language-model critic and clinical expert review complete-panel reasoning and tool-use trajectories to localize recurrent failures and form candidate revisions; the expert may reinterpret the evidence and retains authority over admission and rollback, while Recall@1 and Recall@5 validate outcomes after candidate execution. Across cumulative ablations and ultra-rare-disease benchmarks, a PIHF-derived policy improved Recall@1 in one proprietary executor and three open-weight executors spanning 3 to 49 billion active parameters. Gains were 32.7 percentage points for GPT-5.4 and 31.1 points for Qwen3.6-35B, a difference of 1.7 points. These results support the feasibility of using pretrained language models as fixed-weight execution substrates for expert-guided policy development in rare-disease diagnosis.

The Central Place Foraging Algorithm (CPFA) combines site fidelity, pheromone-guided navigation, and uninformed random search to enable decentralized resource collection in robot swarms. However, CPFA often revisits previously explored regions while leaving other areas insufficiently searched, reducing efficiency as resources become scarce. In this paper, we propose Adaptive Repulsive Pheromone Clustering (ARPC), a bio-inspired method in which robots deposit repulsive pheromone waypoints to mark previously explored locations. These waypoints are clustered around the nest to estimate low-value search regions, allowing robots to be redirected toward likely unvisited areas. By integrating the exploitation of known resources with systematic avoidance of redundant exploration, ARPC improves search diversity and resource discovery efficiency. Extensive simulations in ARGoS across varying arena sizes, resource densities, and clustered, random, and power-law spatial distributions demonstrate that ARPC consistently outperforms CPFA and the Grid-Based CPFA (GPFA). In particular, ARPC yields significant gains during both early discovery (10\%) and late-stage (up to 60\%) collection, where conventional methods typically degrade. These results indicate that ARPC provides a scalable and robust strategy for large-scale heterogeneous swarm foraging environments.

Optoelectronic resistive random-access memory elements (ORRAM) are critical emerging devices that leverage photonic technologies to bring the advantages of optical programming to traditionally electronic memristive platforms for neuromorphic computing and artificial intelligence. In this work, a free-space optic micro-LED ($μ$LED) array is combined with a 2-terminal oxide semiconductor ORRAM (based on IGZO\textsubscript{Rich}/IGZO active layers), to realise parallel spatial programming of form-free memristive arrays. We report the optical and electrical programming of resistive states with potentiation/depression analysis of various stimuli parameters (pulse frequency, pulse width, pulse amplitude). Further, we demonstrate the simultaneous photonic-electronic programming of the ORRAM with optical SET (blue 450\,nm) and electrical RESET functionality. Persistent photocurrent is also observed and exploited as a pathway to fading memory or synaptic plasticity for temporal bit encoding. Finally, parallel optical $μ$LED to ORRAM channels are demonstrated to achieve the simultaneous photonic programming of multiple devices and the writing of spatial patterns across a chip of memristive IGZO devices. This work highlights the light-enabled scalability of the optoelectronic platform and the feasibility of ORRAM to interface with spatially-multiplexed optical sources to bring neuromorphic technologies directly into applications that process and sense in the optical domain.

Large Language Models (LLMs) have demonstrated capabilities in in-context learning, task decomposition, step-by-step reasoning, and code generation, driving their gradual evolution from text generation models into the core of agents capable of perceiving environments, invoking tools, and executing tasks. Traditional LLM Agents typically obtain information through webpages, documents, databases, or external tools and generate corresponding invocation sequences according to user goals; when this technology is further integrated with robotic systems, large language models begin to undertake functions such as task understanding, high-level planning, and behavioral decision-making. SayCan combines the task reasoning capability of language models with the affordances of robotic skills, while Code as Policies and ProgPrompt generate robot task plans through policy code and programmatic prompting, respectively, and VoxPoser uses language models and vision-language models to construct three-dimensional value maps to guide robotic manipulation \cite{6,7,8,9}. Vision-language-action models such as PaLM-E, RT-2, and GR00T N1 further strengthen the connection among language, visual perception, and robotic actions \cite{10,11,12}. In such LLM-driven embodied agents, the model not only needs to understand user instructions, but also needs to combine scene states, object attributes, spatial relations, and execution feedback to complete task grounding, and then hand the generated action plan to skill libraries, motion planners, or controllers for execution.

Agent harnesses have substantially improved performance on long-horizon tasks by coordinating agent interactions with the environment. However, reinforcement learning through complex harnesses remains largely unexplored, as scaling such training to long-horizon agent tasks introduces fundamental challenges. In this work, we present a unified black-box RL framework for stable and scalable optimization of general agents through complex harnesses. Concretely, we first build a sandbox-based execution infrastructure that isolates task environments and harnesses within temporary sandboxes for large-scale concurrent rollouts. We then decouple policy optimization from opaque harness execution and place a serving proxy at the model boundary to capture model calls. To reconstruct multi-turn trajectories and improve training efficiency, we organize the captured calls into prefix trees and further adapt both critic-based PPO and critic-free GRPO to optimize over the recovered tree structure. Meanwhile, we maintain training-inference consistency throughout the optimization process. Finally, we introduce mix-harness training, allowing a single model to be jointly optimized by heterogeneous harnesses. With Qwen3-30A3B, black-box RL improves Pass@1 on ClawGym-Bench by 9.98 and 14.81 points through OpenClaw and Claude Code, respectively, while remaining stable over 200-400 optimization steps. Moreover, the framework yields consistent gains on more challenging tasks such as JobBench and OfficeQA. Overall, our framework enables effective, stable, and scalable optimization of general agents through black-box harnesses, supporting unified training across heterogeneous execution systems.

Language and vision-language models generate plausible embodied plans but do not guarantee executability, as their outputs can violate environment dynamics or act on incorrectly grounded entities. We present a neurosymbolic agent that factors long-horizon household tasks into task-directed visual exploration and constrained symbolic planning. In the first phase, a vision-language model and exploration harness acquire goal-relevant predicates and instance bindings from egocentric observations and grounded interactions, producing a symbolic initial state. In the second, a PDDL transition model restricts decoding to tokens that extend applicable actions. Monte Carlo tree search then evaluates executable continuations using a domain-independent planning heuristic. The resulting plans are executable by construction under the transition model, with transfer to the environment conditioned on correct visual grounding. On VirtualHome and ALFWorld, open 4B-27B models exceed 90% success in both environments, and our smallest agent substantially outperforms a 27B direct visual policy in each. Constraints and search prove complementary rather than interchangeable: in ALFWorld either alone solves under a third of tasks, whereas their combination solves over 95%. The method also uses several times fewer generated tokens than extended thinking and far fewer model-visible images than direct interaction, and residual failures localize to state acquisition rather than plan generation without any specialized training.

Long-term manipulation planning requires robots to reason not only about immediate task success but also about how current decisions affect future interactions with the environment. In this context, household service robots may need to organize groceries in partially occupied shelves while using limited storage space efficiently and preserving access for subsequent placements. In this paper, we consider an online multi-object shelf-placement setting in which future objects arrivals are unknown. Existing approaches do not jointly address semantic organization, dense space utilization, and manipulator accessibility during sequential shelf filling. To address this gap, we propose Semantic-Dense Placement Planning (SDPP), an accessibility-preserving approach that ranks candidate poses using a semantic-density score combining inter-object semantic similarity with spatial proximity. An Accessibility Map (AM) further filters candidates unlikely to be reachable before motion planning and penalizes placements that reduce the remaining accessible workspace. Simulation experiments show that SDPP significantly improves semantic placement quality over state-of-the-art baselines and achieves the highest average shelf density, while the AM substantially reduces the time required to identify feasible placement poses. A qualitative real-world experiment demonstrates the applicability of our pipeline in a domestic shelf-storage scenario.

Designing large soft robots capable of generating high forces for physical human-robot interaction remains a significant challenge in soft robotics. Prior work in large soft robots has focused on proof-of-concept prototypes, and no systematic framework exists for determining the suitability of a design paradigm for a desired task. This manuscript introduces a method for optimizing the geometry of a soft robot limb, maximizing its blocking force subject to an anti-bucking constraint under its own gravitational loading. We demonstrate that an explicit solution exists to the proposed optimization problem under certain assumptions. Experiments with three geometries of a large, soft, pneumatically-actuated manipulator demonstrate that the method correctly predicts which designs meet constraints and which produces the largest end-effector forces. This method, with its closed-form solution, can allow designers to determine a-priori if an intended class of soft manipulators is an appropriate choice for physical interaction at large size scales.

In-context imitation learning enables few-shot policy generalization but struggles to maintain performance on unseen objects and novel scenarios. To address this, we introduce MatchingPolicy, a correspondence-driven framework that explicitly decouples demonstration-to-scene matching from policy learning. Central to our method is a correspondence-aware diffusion policy that conditions robotic actions directly on dense semantic correspondences. This architectural separation resolves the inherent conflict between correspondence identification and action adaptation, enabling robust out-of-distribution transfer. Our framework integrates vision foundation models with a novel two-stage matching algorithm to dynamically establish reliable correspondences. Extensive evaluations on RLBench and real-world manipulation tasks confirm that MatchingPolicy achieves superior few-shot performance, generalizing reliably across unseen object instances and semantic categories.

Underactuated tendon-driven hands offer compact actuation and passive compliance, but tendon elongation under restoring-spring loading introduces configuration-dependent joint deviations. This paper presents H-PAC, a modular 6-actuator, 15-DoF robotic hand with a control-oriented modeling and implementation framework. A sparse analytical actuator-joint model is derived from the tendon-routing geometry, and a mechanics-based compensation model is developed to account for tendon-elasticity-induced joint errors. The proposed method is implemented in a hierarchical architecture: a host computer performs workspace-constrained posture mapping and compensation, while an ESP32 generates synchronized commands for six position-controlled servos. The same control parameters and execution strategy are used across all tasks without task-specific retuning. Monotonic servo-sweep experiments show that the compensation substantially improves joint-angle prediction. The MAE of the index DIP joint decreases from 1.15 degrees to 0.18 degrees, and all nine evaluated joints achieve an MAE below 0.23 degrees. Representative postures and grasping configurations are further executed using the same control pipeline without external joint or force sensing in the control loop. The results demonstrate a practical approach to improving posture reproducibility in compact underactuated robotic end-effectors.

Large language models (LLMs) are increasingly deployed as decision-making agents in settings that require sophisticated environmental exploration. However, existing work has raised questions about how LLMs actually balance exploration and exploitation. Unlike classical agents, LLM agents engage with tasks through natural language, exposing them to semantic information with no formal counterpart in the task structure. We introduce the semantic bandit, an extension of the multi-armed bandit setting that explicitly considers the textual labels assigned to actions, and use it to study how semantic priors --- inductive biases arising from associations between language and expected reward learned during pre-training, shape LLM exploration behaviour. We find that semantically informative action labels reduce exploration in favour of exploitation, improving performance when aligned with the reward structure and severely degrading it when misaligned. We further find that negative rewards trigger substantially more exploration than equivalent positive rewards, consistent with an expected-scale bias induced by reward conventions common in pre-training data. Overall, we argue that the use of language to define the environment and rewards introduces unavoidable biases derived from the fact that the model is trained on word co-occurence, with implications for the reliability and robustness of LLM agents in real-world decision-making settings.

Empathetic social robots should respond not only to what users say, but also to how their emotions dynamically evolve during interaction. However, existing empathetic dialogue systems are often text-centered and primarily model empathy as a one-way mapping from the user's emotion to the system response, limiting their ability to capture embodied speaker--listener affective exchange. We present AffectLoop, a multimodal speaker-listener emotion-dynamics-aware spoken dialogue system implemented on the Misty II robot. The system tracks the speaker's verbal and facial affective dynamics, estimates the robot listener's own verbal and behavioral affective state, and conditions LLM-based response generation on both affective streams. The robot then generates a short spoken empathetic response together with emotionally congruent embodied behavior, forming a closed speaker--listener affective loop. We evaluate the system in a pilot within-subject study with five participants, comparing it with an otherwise identical utterance-conditioned baseline that omits the speaker- and listener-affective-state inputs. The proposed system received higher overall impression ratings, especially for empathetic response and user satisfaction. Post-hoc log analysis further showed higher speaker-listener affective alignment and stronger valence-based distress recovery. These preliminary results suggest that explicitly modeling both speaker emotional dynamics and listener affective state can improve embodied empathetic interaction.

Cross-view Video Geo-localization (CVG) aims to localize ground-view videos by retrieving their corresponding geo-tagged aerial images. However, CVG approaches rely on fixed-length inputs and post-hoc refinement, hindering online-oriented localization under partial or dynamic observations. In this work, we formulate Progressive Cross-view Video Geo-localization (PCVG) as a deployment-oriented extension and evaluation protocol of CVG, enabling localization under varying temporal budgets, prefix-based inference, random-start evaluation, and long-range localization with interruptions. To explore PCVG, we introduce X$^2$Localizer, a cross-grained alignment framework that jointly supervises global prefix-to-aerial retrieval and token-aggregated frame--aerial-tile matching with a budget-dependent asymmetric objective. Furthermore, we introduce a Sliding-Window Re-Localization (SWRL) strategy that dynamically refreshes candidate regions for failure recovery and long-range deployment without full-sequence reprocessing. Extensive experiments show that X$^2$Localizer preserves conventional full-video performance, with marginal gains of +0.1 Recall@1 and +0.3 Recall@10, while substantially improving early localization. In the challenging single-frame setting, X$^2$Localizer improves coarse retrieval by +4.7 Recall@1 and +11.5 Recall@10 over the previous state-of-the-art method. With SWRL, our approach further enables robust progressive localization under random-start and long-distance scenarios, narrowing the gap between benchmark evaluation and real-world deployment.

Satellite agents for on-orbit navigation tasks need to predict collision risks using limited onboard observations. However, conventional planners often rely on predefined maps and fixed environmental assumptions, limiting their adaptability in dynamic on-orbit scenarios. In this paper, we propose Orbit-Planner, a two-stage latent world model for on-orbit obstacle avoidance. Orbit-Planner learns action-conditioned spacecraft dynamics to perform future-state rollouts in latent space, and introduces a Physics Probe to decode physical state changes from imagined latent trajectories. Experiments demonstrate that Orbit-Planner can perform long-horizon latent rollouts and recover physical states from imagined trajectories. In closed-loop obstacle-avoidance navigation in Isaac Sim, it attains a success rate of 91.7%. Code is available at https://github.com/ZhijianLi2003/Orbit_Planner.

On-policy distillation (OPD) transfers teacher capabilities by supervising trajectories sampled from the student's own policy, yet its generalization behavior remains poorly understood, as most studies evaluate OPD on a single domain and on benchmarks close to the training data. We present a controlled study that varies one generalization factor at a time, from in-domain distribution shifts to cross-domain transfer and the multi-teacher setting. We find that OPD transfers a teacher's reasoning behavior rather than its answers to particular problems: training difficulty barely matters, and even problems the teacher never solves are useful. Transfer depends strongly on the origin relationship between teacher and student: same-origin pairs bring the student close to the teacher across languages, reasoning horizons, and even other domains, whereas cross-origin pairs mostly fit the trained distribution. This broad reach is a double-edged sword: since routing prompts to domain experts cannot confine each teacher's influence, combining them yields a mixture-dependent seesaw among their capabilities. These results clarify when OPD generalizes and offer a useful perspective for diagnosing multi-teacher OPD.

Can a language model recover the true research idea of a published paper when given only that paper's pre-publication bibliography? We introduce Reconstruction, a blind idea-recovery benchmark that withholds the seed paper and all contemporaneous or future literature, and asks models to propose hypotheses that an independent large language model judge matches against the held-out ground-truth idea. A strict anti-leakage protocol-temporal citation cutoff, anonymous reference IDs, and frozen per-paper bibliographies, which prevents prompt-time leakage of the seed idea. Across six scientific domains and 643 evaluated papers, seven frontier models achieve only modest Match rates (approx. 3-15%). We then evaluate a reference-only multi-agent (top 4) pipeline that combines cross-model review with a Swiss tournament over aligned hypothesis slots, without external web search. Cross-model review plus tournament selection raises Match rates to approx. 23-42% across all six domains, which is an observed approx. 2.4x lift over the best single-model baseline. This draft reports the protocol, anti-leakage design, and current results as an arXiv timestamp.

Automated detection of errors in clinical documentation is a promising application of large language models (LLMs), yet decisions to deploy such models rest on benchmarks that evaluate each clinical note in isolation. Error-detection benchmarks are typically constructed by injecting errors into notes, such that each erroneous note has a natural counterpart. Aggregate discriminative metrics (e.g., balanced accuracy or F1) do not exploit this structure. We show that this omission is consequential. In particular, evaluating 15 diverse LLMs on 4 standardized clinical error-detection test sets across 3 languages, we find that 13 of 15 models fall below the level of random pairwise discrimination, even while achieving F1 scores that standard practice would read as moderate. We also observe that the underlying bias patterns differ across languages: the same model can default to "no error" on one language and over-flag errors on another. To diagnose where discrimination breaks down, we further introduce a procedure to score the evidence models cite in their outputs. We find that while models consistently locate error-relevant content, they fail to produce the corresponding correct verdict on the clean counterpart. Finally, we show that F1 and pairwise accuracy are driven in opposite directions by the same underlying bias, so that ranking models by F1 may systematically promote the weakest discriminators. For safety-critical clinical NLP applications, we advocate for supplementing aggregate metrics with paired evaluations in benchmark reporting. Code and analysis scripts are available at https://github.com/healthylaife/paired-clinical-eval.

Accurate throwing of the American football requires precise regulation of release conditions, where coupled linear and angular momentum determine flight stability and targeting accuracy. While prior work on robotic object throwing has largely focused on generating dynamically feasible release velocities using open-gripper paradigms, explicit control of spin injection at detachment remains underexplored, particularly for aerodynamically anisotropic objects like the American football. In this paper, we present the spin-stabilized controlled tight spiral throw of an American football by a humanoid robot. Achieving this requires (i) accurately reaching the desired coupled momentum, which often involves high degrees-of-freedom (DoF) movements completed within approximately half a second, and (ii) managing the complex transient contact dynamics that arise during the sub-100-millisecond release phase, when the football is effectively underactuated as it moves partially across the fingers. To this end, we develop a coupled whole-body control strategy where the lower body is performing informed stabilization while the upper body is further divided into two phases with (i) a throw phase accelerating the football to a target state through trajectory optimization and tracking, and (ii) a follow-through phase utilizing model predictive control to actively control the wrist and remaining in-contact fingers. The proposed framework is empirically validated on a 29-DoF Unitree G1 humanoid equipped with a 7-DoF Dex3-1 three-fingered gripper. The thrown American football reaches up to 93.6% spin efficiency and a 0.286 radians linear-velocity-to-nose-alignment (nose-angle) error (where an ``ideal'' tight spiral corresponds to 100 % spin efficiency and 0 radians nose-angle error) at up to a 5.35 m/s linear velocity and an angular velocity of 14.5 rad/s.

Vision-based Unmanned Aerial Vehicles (UAVs) often suffer from navigation failures in dead ends due to limited sensing accuracy and range. To address this challenge, this paper proposes a systematic solution for efficient dead-end prediction and avoidance. The proposed method introduces a lightweight neural network to predict the relative distance and bearing of potential dead ends within the current field of view using RGB-D inputs. These predictions prune a predefined, compact trajectory library, enabling the planner to proactively avoid dead ends while maintaining navigational smoothness. Notably, our approach transfers across real-world scenarios without manual annotation or fine-tuning on real-world data. The system achieves high-frequency replanning at 50 Hz onboard. Extensive simulation benchmarks demonstrate superior performance in success rate, flight time, and trajectory length, and real-world experiments further validate its effectiveness in complex scenarios.

Natural language explanations (NLEs) are increasingly used as inputs, for example, as few-shot rationales that influence model behavior in in-context learning (ICL). However, it remains unclear how different types of NLEs compare in their effects on downstream model performance in explanation-augmented prompting. Therefore, we provide a comparative evaluation across six benchmarks and four instruction-tuned models, studying how NLE source (human-written when available, self-generated explanations, generated by an external LLM) and NLE selection (random vs faithfulness-based filtering) affect downstream utility of NLEs when used in ICL settings. Our extensive evaluation shows that, on classification-style benchmarks, adding NLEs to few-shot prompts often improves accuracy over few-shot prompting without explanations; among NLE sources, externally generated LLM-NLEs often provide strong downstream utility and remain competitive with human rationales where both are available, whereas self-NLEs are more sensitive to the selection strategy. On math reasoning, the effects are more model- and source-dependent. We further show that faithfulness-based selection of self-NLEs yields small average gains overall, but can improve or reduce performance depending on the metric, task, and model. Different faithfulness metrics can disagree substantially, affecting which self-NLE examples are selected and their downstream predictive utility. Robustness tests with randomly swapped and out-of-distribution rationales indicate partial robustness, suggesting that semantic alignment contributes to performance gains. Overall, our results provide insights for selecting and reporting explanations that influence model behavior in practical prompting pipelines.

Palmyra x6 is a large language model optimized for use with enterprise-oriented agentic tasks. The model was built by post-training a Mixture-of-Experts base model with Anchored Supervised Fine-Tuning on a compact corpus of verified, synthetic tool-use trajectories, optimized with a Muon + Adam hybrid. The recipe is deliberately conservative and deliberately controlled: 626 trajectories, a single epoch, a low learning rate, and a KL anchor to the frozen base. The model shows substantial gains over the previous default model for Writer Agent, and compares favorably with several recent models on public benchmarks, scoring the highest on BFCL Core at $0.785$ and posts the highest six-benchmark mean of the cohort. Furthermore, the model has shown itself to be competitive or leading relative to comparators in our bias and safety evaluations.

High-resolution three-dimensional (3D) nanostructures for visible-light photon manipulation provide unique and bespoke capabilities in optics and photonics. However subwavelength nanofabrication and reliable ensemble manipulation of the 3D prints onto arbitrary substrates remain challenging. Here, we introduce sublime transfer strategy tailored for transfer printing ensembles of delicate 3D printed nanostructures. This strategy enables conformal, damage-free integration of arrays of 3D structures on diverse substrates. Naphthalene acts as a transient stamp to encapsulate the structures during transfer and placement. We rely on the low sublimation temperature of naphthalene to release the structures reliably with nearly zero stress, preventing mechanical damage and positional misalignment. This approach is broadly applicable to integrate diverse nanostructures and photonic devices onto various substrates, and enabling inorganic architectures through ensemble uniform post-processing, including 2.5D photonic crystals on flexible PDMS, diffractive optical elements on curved lenses, spiral phase plates on CMOS chips, multilayer achromatic metalens on optical fiber facet, as well as 3D glass photonic crystals and optical topological resonators on anti-stiction quartz.

Embodied agents are increasingly used to close the gap left by end-to-end policy models. Yet the agentic path has not realized closed-loop learning in physical execution: existing harnesses remain largely open-loop, following fixed skills during rollout and reflecting only after an episode completes. Such post-hoc reflection cannot govern execution as it unfolds, because physical interaction requires decisions to track rapidly changing robot-environment states at a frequency beyond today's large agentic models. We present Zetta, a closed-loop embodied harness that evolves code-based runtime critics and recovery skills online while keeping the base policy frozen. Through three timescale-separated loops, Zetta provides action-frequency governance, rollout-level critic-recovery proposal, and validation-gated skill updates. Together with Z-Infra, a rollout infrastructure decoupling agent logic from heterogeneous execution resources, Zetta achieves state-of-the-art success on LIBERO-Pro and RoboCasa under our current rollout budget, reaching 90.8% and 93.6%, with an 11.1x inference speedup; success continues to scale with self-exploration experience; learned skills transfer zero-shot, and clear robotic "Aha Moments" emerge. These results show that closed-loop harness self-evolution opens a scaling path for reliable physical intelligence.

AI agents increasingly operate as part of interacting systems rather than in isolation. As agents exchange information and jointly make decisions, their interactions can improve collective reasoning but may also produce herding, polarization, or amplify shared biases. Understanding and predicting these collective dynamics is therefore important for designing effective and aligned multi-agent systems. Here, we study over 10,000 communities of language-model agents that repeatedly exchange messages and revise their opinions across objective mathematics questions and subjective political statements. Despite substantial diversity in possible behavior, the individual and group dynamics can be represented by three characteristic regimes: indifference, polarization, and consensus. AI agents start indifferent and build conviction as they interact. On objective questions, communication improves collective accuracy, while on subjective questions it often drifts group opinions toward the right in the political spectrum. We explain these observations with a statistical-mechanics formalism in which agents stochastically favor lower social pressure. Given only initial opinions, our model predicts individual trajectories, outperforms all standard baselines, generalizes to unseen community graphs, and reproduces the observed group archetype distributions. Our fitted model parameters reveal the mechanics underlying our key observations: i) communities operate below the critical social temperature, which explains conviction buildup; ii) attractive ties outweigh repulsive ones, which favors consensus; and iii) agents holding the correct answer exert the strongest pull, which drives truth-seeking. Overall, our results demonstrate that collective behavior of AI agents, like that of other complex systems, follows compact and predictive dynamical laws.

Learning from demonstration is a promising approach for dexterous manipulation, but collecting high-quality contact-critical demonstrations remains difficult with low-cost teleoperation hardware. We present ViHaTeleop, a lightweight (0.7 kg), low-cost (\$550) visual-haptic teleoperation system with SLAM-based wrist tracking, camera-based hand tracking, and finger-wise vibrotactile feedback through Linear Resonant Actuators (LRA). The system includes several design choices (LED illumination, fisheye hand camera, and tactile-aware retargeting constraints) and is deployed on Franka + LEAP Hand + 9DTact in both real and simulated environments. Under matched with/without-haptic conditions with nine participants across six contact-critical tasks, haptics improved success rates across all tasks (+2.2 to +15.6 percentage points), while completion-time effects were task-dependent. Subjective ratings showed significant gains in contact clarity and grasp confidence in both simulation and real-world settings (Wilcoxon signed-rank, $p<0.05$). We also integrate a lightweight depth-camera-based tactile proxy in Isaac Sim, enabling a full pipeline from multi-modal demonstration collection to visual-tactile policy training. Preliminary downstream validation by training visual-tactile policies from collected demonstrations shows tactile cues benefit contact-critical subtasks (peg-in-hole: +17 percentage points over vision-only).

Recent advances in AI technologies have enabled the advanced design of complex control policies. In contrast, focusing on the body, many robots still employ simple bodies that can limit adaptability to environments. Studies in embodied robotics have shown that well-designed bodies can partially replace the role of control and computation with physical body-environment interactions, yet such designs still depend heavily on expert intuition. There is a need for a systematic theoretical framework for body design, as well as a method for joint optimization of the body and controller. To address this, we introduce the Embodied Perceptron, a theoretical framework that unifies neural networks and physical body systems. In this view, the body itself acts as a perceptron: mechanical parameters correspond to weights, and physical nonlinearities play the role of activation functions. By representing physical constraints as weights and nonlinear properties as activation functions, a physical body can be modeled in neural-network form. The system representation enables us to explicitly and theoretically explain that the body can substitute for part of the neural control. As an application, we co-optimize control policy and muscle configuration in a musculoskeletal robot and show that the resulting embodied intelligence can provide inherent stability, improve learning efficiency, and drastically reduce model size-even with a single-neuron controller. The results bridge the informational and physical worlds and provide a pathway toward understanding and systematic design of embodied AI systems.

Answer-only reinforcement learning (RL) trains reasoning models to solve fully specified problems, but many realistic queries omit a premise needed for a unique answer. In this setting, the useful response is not always refusal: the model should ask for the missing premise, condition its answer on the unknown quantity, or abstain when no informative conditional response is available. We present \emph{Ask-Condition-Abstain Reinforcement Learning} (ACA-RL), a data-augmented RL framework for this setting. Its reasoning-graph-guided pipeline converts well-posed problems into missing-premise training instances with localized gap annotations; ACA-RL then trains on these instances with a structured reward over five observable response behaviors. We also introduce the \emph{Missing-Premise Benchmark} (MPB), a 274-instance human-verified benchmark spanning mathematical, logical, and real-world word problems. Across Qwen3 and Llama models, ACA-RL consistently improves on MPB while preserving competitive performance on well-posed reasoning tasks. Together with the released code, MPB, and training data, this work supports a new mission for NLP evaluation: measuring whether models can recognize when a task is underdetermined and handle uncertainty, not only whether they can answer fully specified questions.

Multi-preference alignment is often framed as scalarization: combine reward dimensions, then optimize. This leaves a temporal decision underspecified: when should each preference dimension enter policy optimization? We propose \methodname, a stability-guided active-set controller for controlled objective admission. \methodname starts from a small active set, retains admitted objectives, and expands when reward-deviation gates indicate low recent deviation or a patience budget is exhausted. A probing phase estimates a hard-to-easy order, and adaptive weighting emphasizes underperforming active dimensions. Automatic evaluations with 15 training preferences and 16 held-out benchmark columns show that \methodname obtains higher averages than simultaneous scalarization and shared-budget adapted baselines. Component ablations and expansion dynamics further support cumulative retention, gated admission, and probing-derived ordering as useful design choices in this setting. These results position objective-entry timing as a concrete control variable in reward-vector RLHF.

Agents increasingly rely on reusable skills to encode task knowledge, tool-use procedures, and validation rules. Existing skill self-evolution methods primarily revise skills using execution trajectories collected from current tasks, leaving the evolution knowledge accumulated in public skill version histories largely untapped. Our pilot study reveals a clear complementarity between the two sources: public skill changes provide reusable evolution priors, whereas trajectories provide evidence grounded in the current task. Motivated by this, we propose VCE-Skill, which distills noisy and implementation-specific public skill changes into reusable, structured version-change experience and adaptively fuses it with trajectory-derived proposals from the base evolver, thereby exploiting external experience while retaining task-specific evidence. Extensive experiments demonstrate that VCE-Skill improves skill self-evolution, increasing mean scores by 3.20--4.98 points; transfer experiments further show that the resulting skills achieve stronger cross-model transfer performance. Our work highlights public skill version changes as a previously underexplored yet effective source of prior knowledge and advances trajectory-driven skill self-evolution.

Large Audio Language Models (LALMs) have made rapid progress on standardized benchmarks, yet their deployment in practical media workflows, curation, archival indexing, and content distribution remains largely unrealized. We identify automated audio chapterization, the task of segmenting continuous audio streams into thematically coherent chapters, as a demanding and commercially consequential setting that exposes this gap. Chapterization is challenging because boundaries are defined less by objective acoustic events than by subjective editorial judgment, requiring models to reason sequentially over long acoustic contexts and approximate creator-authored boundary decisions. We present AudioChaps, a post-training framework for aligning end-to-end LALMs for this task via Group Relative Policy Optimization (GRPO) guided by Chain-of-Thought (CoT) reasoning. To support training and evaluation, we curate three datasets: AudioChaps-Alignment, derived from creator-annotated chapter boundaries on YouTube; AudioChaps-CoT, which provides structured supervision for well-formatted, high-quality, and evidence-grounded boundary reasoning; and AudioChaps-Eval, a held-out benchmark for audio chapterization. Applying GRPO directly without a Supervised Fine-Tuning (SFT) cold start, AudioChaps-R1-Zero already improves average F1 by 33 points over the state-of-the-art LALM Audio-Flamingo-3-Think. The AudioChaps framework produces our final aligned LALM, AudioChaps-R1, which improves average F1 by 49 points. These results demonstrate that GRPO-trained LALMs can reliably transform unstructured auditory streams into navigable, structured media. Our code, models, and dataset resources will be released upon acceptance at https://github.com/ta012/AudioChaps.

Multimodal Retrieval Augmented Generation (M-RAG) is increasingly vulnerable to adversarial attacks where malicious data are crafted to produce embeddings that align with benign entries in the vector space, deceiving retrieval and inducing harmful outputs. Existing defenses primarily operate at query time, relying on auxiliary detectors, similarity re-ranking, or feature-consistency checks. However, these approaches suffer from non-trivial inference overhead, generalize poorly to unseen attack strategies, and often assume specific attack distributions. To address this, we propose DSPrompt, a Dynamic Soft Prompt defense framework that directly reshapes the retriever's embedding semantics, without modifying the retrieval pipeline. It inserts few learnable soft prompts into each layer of the visual and textual encoders of a frozen retriever, utilizing a shallow-to-deep length schedule that is adaptive to the capacity in the model layers. These prompts are trained under a dynamic min-max scheme: an online multimodal attacker continually crafts hard adversarial documents against the current retriever, while the defender is updated to push such documents out of the top-k while preserving the ranking and diversity of benign evidence. Because the defended encoder can be pre-computed and indexed exactly as in standard dense retrieval, DSPrompt incurs no additional per-query optimization and introduces fewer than 1% additional parameters. Extensive experiments across four benchmarks and three representative poisoning attacks show that DSPrompt substantially reduces the attack success rate and poison retrieval rate while maintaining near-lossless retrieval utility and generation fidelity, consistently outperforming existing defense baselines at a fraction of their computational cost.

Retrieval-augmented generation (RAG) improves large language models by grounding generation in external evidence, but it also introduces a source trust problem: retrieved context may be useful, irrelevant, or even misleading. Existing RAG systems often apply a fixed trust policy toward retrieved evidence, which can either over-trust incorrect context or underuse context when the user explicitly asks for context-following behavior. Therefore, we propose Intent-Guided Decoding (IGD), a framework that arbitrates between retrieved context and parametric memory according to user intent. IGD uses answer-level filtering and token-level correction to steer the final decoding trajectory between retrieved context and parametric memory. We evaluate IGD on three faithful QA benchmarks and three factual-conflict benchmarks across five LLMs, IGD substantially improves factual recovery, achieving gains of up to 65.4 percentage points on factual-conflict benchmarks over Direct RAG, while preserving or improving strict context-following behavior, this findings highlight the importance of balancing factuality and faithfulness in RAG.

Real-world deployment of Vision-Language-Action (VLA) models is often bottlenecked by efficiency-performance trade-offs, cross-embodiment generalization, and execution smoothness. We present NebulaVLA, an asynchronous dual-frequency architecture that decouples high-level semantic reasoning from low-level action control, optimizing computational resources and modularity. To bridge semantic gaps across heterogeneous robots, we introduce GESTURE-7, a unified language-grounded action representation. Furthermore, our Guide Action algorithm enforces kinematic continuity via mask-based smoothness constraints. Comprehensive evaluations demonstrate that NebulaVLA significantly outperforms synchronous baselines, achieving an 85.5\% average success rate on LIBERO-Plus and accelerating action generation by \textasciitilde 2.7$\times$. This asynchronous design enables highly efficient and responsive control for practical robotics.

Active 3D Gaussian reconstruction fundamentally relies on selecting informative next-best views under limited sensing budgets. Existing active 3DGS methods primarily plan viewpoints according to geometric information gain, treating object-induced hidden regions in the same manner as general unexplored space. Under tight frame budgets, such geometry-driven strategies may prioritize global scene coverage while leaving partially observed objects incompletely reconstructed. To address this limitation, we propose OccamView, an object-conditioned view-selection framework for frame-budgeted active 3D Gaussian reconstruction. Rather than predicting unseen object geometry or performing shape completion, OccamView maintains an online object memory from open-vocabulary detections grounded in measured RGB-D observations and represents unresolved local occupancy around detected objects as conservative hidden-region proxies. Candidate viewpoints are then evaluated using an occlusion-aware proxy-coverage score. Furthermore, we introduce a Geo-Floor mechanism that restricts object-conditioned re-ranking to geometrically competitive candidates, allowing object-conditioned cues to guide complementary observations while preserving the geometry-driven exploration behavior of the underlying planner. Experiments on Replica and Matterport3D under a unified frame-budgeted protocol show that OccamView consistently reduces Completion and improves Completion Ratio across five frame budgets, with particularly pronounced gains under limited frame budgets. These results demonstrate that lightweight object-conditioned cues effectively complement geometry-driven active view planning.

Modern power systems face growing operational complexity driven by the integration of renewable energy sources, decentralization, and the need for real-time decision-making across a wide range of timescales. Addressing these challenges traditionally relies on model-based methods that, while accurate, can be too slow for operational demands. Machine learning (ML) has therefore emerged as a faster, data-driven alternative. As grid topology plays a central role in power system operation, graph machine learning (GML) methods offer a natural framework for incorporating topological dependencies as an inductive bias. We survey nearly 800 papers at the intersection of GML and power systems, covering forecasting, state estimation, optimization, control, fault diagnosis, and cybersecurity. Power systems constitute an unusually rich benchmark setting for GML, as they combine hard physical constraints, multi-scale dynamics, safety-critical requirements, and scarce labeled data within a single, well-defined domain. Conversely, power systems can benefit from utilizing GML to complement classical solvers, as GML provide scalable, topology-aware approximations with promising generalization and computational efficiency. We identify open challenges, including limited real-world deployment and the need for interpretable models in safety-critical settings. Despite the rapidly growing number of publications, standardized benchmarks and open datasets remain scarce, leaving many results difficult to reproduce and undermining the long-term scientific credibility of the field. We further derive a structured requirements catalog for ML-ready power grid benchmarks, intended to guide future dataset development and improve reproducibility across studies. We call on the community to prioritize dedicated benchmark studies and the release of open datasets and models.

Learning embodied urban navigation policies from real-world data is constrained by the cost of task-specific data collection and the limited coverage of rare yet safety-critical scenarios. To address these challenges, we present a scalable framework for learning point-goal urban navigation from web-scale in-the-wild egocentric videos while systematically exposing its long tail. The framework automatically annotates uncurated web videos with metric trajectories and structured navigation semantics, which are then used to train a vision-language-action policy for interpretable navigation planning. We characterize the long tail based on model performance and the distribution of perception-motion patterns, and employ reflection-based analysis to diagnose recurring failure modes. Experiments on web-video data and real-world urban navigation tasks demonstrate effective knowledge transfer from unconstrained videos and reveal coherent long-tail structures beyond aggregate navigation performance.

Long-horizon embodied tasks require LLM agents to iteratively decompose high-level goals, revise plans in response to environmental feedback, and ground leaf-level subgoals into valid executable actions. Recursive context-management methods such as ReCAP improve planning stability through multi-level task decomposition and parent-node refinement, but still repeatedly invoke the LLM at leaf nodes to ground atomic subtasks into exact valid actions. We refer to this final grounding step as last-mile grounding redundancy, which accumulates into substantial LLM-call and token overhead during long-horizon execution. To mitigate this issue, we propose HaReCAP (Habitual-action Grounded ReCAP), a low-intrusion leaf grounding extension for ReCAP. HaReCAP extracts frequent leaf decisions from successful trajectories and compiles them offline into auditable and abstainable one-step leaf-reflex rules. At runtime, it skips the leaf LLM call only when a rule can uniquely determine a legal action in the current valid-action set; otherwise, it falls back to the original ReCAP. This design avoids repeatedly carrying the full recursive context into the LLM for routine leaf action grounding, while preserving the original recursive control flow. We evaluate HaReCAP on Robotouille and ALFWorld with Qwen3.5-27B as the main model. On tasks solved by both ReCAP and HaReCAP, HaReCAP reduces token consumption by 14.67%, 17.93%, and 20.08% on Robotouille synchronous, Robotouille asynchronous, and ALFWorld, respectively. The results show that HaReCAP can serve as a low-intrusion extension to ReCAP-style recursive context-management frameworks, reducing last-mile grounding redundancy across environments and models on commonly successful trajectories.

Inspection is a core capability in many mobile robotics applications, including industrial facility monitoring, infrastructure maintenance, agriculture, and search and rescue. Observing the bottom of a cylindrical cavity, as required by ASTM search-task benchmarks for response robots, presents a representative challenge: the robot must position its camera precisely while satisfying visibility, kinematic, and collision constraints. This paper presents a fully autonomous method for observation-constrained inspection of cylindrical cavities in robot joint space. Rather than prescribing a single Cartesian camera pose, the method represents the inspection objective as a set of valid viewing geometries, thereby avoiding the rejection of reachable viewpoints and configurations with poor joint-limit margins. An RGB perception front end estimates the opening center and directed cavity axis from semantic masks using arc-supported ellipse fitting together with body and side-generator cues. These estimates parameterize constraints on camera-axis alignment, lateral offset, and axial standoff. A multistart derivative-free search then optimizes robot joint configurations with lexicographic priority given to constraint satisfaction; feasible configurations are ranked according to motion economy, joint-limit margin, and view quality. The resulting candidates are evaluated by a collision-aware motion planner, and the executed camera pose is verified geometrically and using a ray-based estimate of bottom visibility. In Isaac Sim, the proposed method successfully completes 92 of 100 target configurations and attains 91.65% mean bottom visibility among executed trials, compared with 76 of 100 and 84.3% for a multistart coordinate-search baseline. Tabletop and Unitree A2-mounted experiments demonstrate the complete perception-planning-execution pipeline.

Multi-modal retrieval-augmented generation (RAG) is a key technique for visually rich long document understanding. Existing multi-modal RAG methods are progressively advancing toward multi-agent systems: they first retrieve relevant pages based on a query, and then iteratively understand information within those pages. However, these methods typically rely on fixed workflows and lack the ability to dynamically scale computation at test time, often leading to insufficient evidence. To address this, we propose D2-ScaleAgent, an agentic framework that introduces a dual-dimensional scaling paradigm for retrieval and reasoning. The core of D2-ScaleAgent is a Verifier agent-driven dynamic routing loop based on the intrinsic difficulty of the query, centered around a continuously updated evidence bank that serves as the agent's dynamic working memory: when retrieval needs to be expanded, the agent routes outward (retrieval scaling), decomposing the query into attributes and performing parallel page retrieval, followed by adaptive pruning to ensure comprehensive evidence coverage. When fine-grained reasoning is required, the agent routes inward (reasoning scaling), dynamically selecting sub-agents with varying granularity and count to extract evidence from pages. Finally, D2-ScaleAgent achieves logical closure over the evidence chain. Extensive experiments demonstrate that D2-ScaleAgent is effective on long and visually rich document benchmarks like MMLongBench-Doc, LongDocURL, etc.

Financial agents must do more than recall domain knowledge: they must be both reliable, executing precise operations over grounded evidence, and executive, sustaining long-horizon research whose conclusions remain auditable. We present Mint-Agent, a family of finance-native agentic models designed around these two scales of financial intelligence. Mint-Agent is built upon three pillars: data, harness, and algorithm. Our data engine constructs clean, specialized tasks for atomic financial capabilities and long-horizon agentic execution from real-world financial sources. MintHarness enables stable interaction with open-ended environments and maintains auditable evidence trails across extended research trajectories. Our training recipe combines SFT, critical-step OPD, and RLVR to develop separate financial reasoning and agentic execution experts, which are then unified through model merging and multi-teacher on-policy distillation into compact, general-purpose financial agents. This pipeline yields two flagship models, Mint-Cu (9B) and Mint-Ag (27B). Across professional financial benchmarks, our models demonstrate two defining strengths: (1) Reliability: Mint-Ag achieves 98.33% on RFC-Bench, surpassing GPT-5.6-Sol and Claude-Opus-4.8 by 3.66 and 3.00 points; and (2) Executability: Mint-Cu reaches 69.86% on FinSearchComp T2, outperforming Agents-A1-35B and Nex-N2-mini by 22.83 and 12.78 points, while Mint-Ag achieves 76.00% and 60.49% on FinanceAgentBench v1.1 and v2, respectively. These results establish a path toward trustworthy financial intelligence in which domain expertise, long-horizon execution, and auditable evidence are jointly engineered as a unified foundation for frontier agentic models.

Autonomous agents share a transport and can call each other's tools, but they cannot share what they know: no protocol lets two agents' memories reconcile a fact phrased two ways, link related facts held apart, or reconcile contradictory knowledge without silently discarding either claim. We present MELD, a self-managing coherence mechanism for a federation of agent memories whose run-time model is the knowledge graph itself. Each brain admits every incoming claim through a five-outcome procedure (insert, merge, relate, conflict, or reject), decided from three signals (scoped claim-key identity, embedding similarity, and a natural-language-inference verdict) under context and freshness gates, and acting through exactly one auditable, authenticated Patch, the only object that mutates state. A binding onto standard publish/subscribe transport with a per-claim status CRDT keeps sovereign brains coherent in claim status without a coordinator: self-healing after partitions and under lossy routing, and self-protecting against silent rewrite by a peer, under a benign-fault model. MELD does not adjudicate truth; a detected contradiction is preserved for later adjudication, never silently resolved. On HotpotQA distractor, distributed merge is recall-non-inferior to a centralized store under a pre-specified equivalence test and recall-superior to naive union at about 11% less live storage; the merge classifier separates at AUC 0.968 with a 0.013 false-merge rate on adjudicated candidate pairs; the status CRDT reconverges in 30/30 real partition-heal trials where last-writer-wins manages 11/30; and semantic routing delivers about 3x fewer messages at matched recall. We evaluate on a real computing continuum spanning an operator-grade 5G edge, national HPC, and a local tier, with empirically calibrated thresholds.

Even well-aligned large language models confidently generate factually incorrect text, making hallucination a persistent reliability risk in high-stakes deployments. These models nonetheless carry linearly separable truthfulness signals in their internal representations. Existing white-box detectors, however, collapse this evidence to isolated components or a single depth, discarding discriminative information distributed across the full forward pass. We introduce HalluTracer, a detection framework that reads and aggregates truthfulness evidence across every layer of the forward pass before the model emits any answer token. A geometric analysis reveals that the per-layer signals are weakly correlated, so that simple depth averaging suppresses layer-specific noise and captures nearly all linearly accessible information. Across six open-source language models and five hallucination benchmarks, HalluTracer consistently outperforms matched white-box baselines, with gains ranging from one to fourteen points. Collectively, our work recasts hallucination detection from a layer-selection problem into a depth-aggregation problem governed by the geometric sparsity of the truthfulness signal.

Dexterous grasp generation that considers arm-related constraints is crucial in real-world scenarios involving arm environment collision avoidance, workspace boundary grasps, and consecutive grasping. Existing hand-centric grasp models, which primarily focus on the floating hand's pose, are insufficient for such cases. Conventional arm-aware methods either rely on rejection sampling to discard infeasible samples or require retraining on arm-specific data, leading to low sample efficiency under adverse conditions or limited generalization across different robots and environments. To overcome these limitations, this letter presents an arm-aware dexterous grasp generation framework that leverages pretrained arm-agnostic grasp models while integrating arm and environmental information only at inference time. Specifically, we formulate arm-aware constrained grasp generation as a joint optimization of hand pose and arm configuration, and derive closed-form gradients for arm-related constraints. Assuming the hand pose distribution is represented by a diffusion model, we prove that gradient-based optimization is equivalent to guided diffusion sampling, steering near-feasible samples toward the feasible region. Through comprehensive evaluation involving 10k objects across 6 scenarios, we demonstrate that the proposed framework generates feasible grasps in highly constrained settings with significantly higher probability, highlighting its advantages in real-world applications. Supplementary materials and appendix are available at https://arm-aware-dexgrasp.github.io/.

Indic quality estimation (QE) and automatic post-editing (APE) data is spread across separate releases, so no single resource supports training and evaluation across tasks and language pairs on one footing. We consolidate the WMT 2020--2024 shared-task lineage with an extended English--Malayalam resource into \indicqe: $126{,}754$ instances over nine directional pairs, with up to four label types aligned on the same segment, a direct assessment, a human post-edit, word-level OK/BAD tags and an error explanation, and a test set stratified over four difficulty axes. On it, we benchmark six prompted LLMs and three COMET metrics on segment-level QE, and three systems on APE. Two of the axes are defined partly on the direct assessment and select a compressed slice of it, so each axis is compared against a control drawn from the same language pair with the same score distribution. Only one survives that control: segments whose holistic and token-level quality signals conflict are ranked worse than equally-scored segments of the same language, for all nine systems and all seven pairs that carry the axis. Annotator disagreement, which looks second-hardest without the control, has no effect with it. Few-shot prompting costs every model $\leq$ $3.4$B both correlation and output-format compliance. Within-language accuracy does not make scores comparable across pairs: of the three trained metrics, the one with the best within-language correlation loses most when the pairs are pooled. The benchmark and code will be released.

On-policy distillation (OPD) aligns a student model with a teacher's logit distribution on student-generated trajectories. This approach has achieved strong empirical gains and can often surpass conventional off-policy distillation with substantially less data. However, standard token-level OPD can provide only fragmented corrections along an erroneous student trajectory and cannot unfold a complete and correct repair path. Motivated by this limitation, we propose \emph{Step-Level On-Policy Distillation} (SOPD), which combines the long-horizon correction of supervised fine-tuning (SFT) with the on-policy advantage of OPD to provide step-level supervision over complete student-generated trajectories. We show that, at different limits of step length, SOPD reduces to SFT or approximates OPD. Compared with SFT, the teacher responses in SOPD are conditioned on student trajectories and therefore align more closely with student-visited states; compared with OPD, SOPD provides longer-horizon corrections rather than fragmented token-level guidance. Across both reasoning and agent tasks, SOPD substantially outperforms conventional SFT and OPD. For example, on ALFWorld, SOPD improves the average success rate by 13.4 points over Vanilla OPD. We hope this work offers a new perspective for future research on distillation methods.

Large Multimodal Models (LMMs) for video reasoning have long been hindered by the high computational cost of processing vast amounts of visual information. This dilemma motivates the transfer of the reasoning capabilities of large models to smaller, more efficient ones. On-Policy Distillation (OPD) offers a promising solution by matching output-token distributions along student-generated trajectories. However, video reasoning often depends on evidence accumulated across multiple frames. In this context, output-level supervision only captures information expressed through token predictions and does not directly constrain the latent representations formed during reasoning. To address this limitation, we propose Latent-OPD, which augments OPD with trajectory-level latent distillation. Specifically, our method focuses on the position at the end of each trajectory, where hidden states effectively summarize the accumulated visual evidence and reasoning context. Furthermore, we introduce a progressive teacher-lookahead strategy, which aligns middle-to-late student layers with increasingly deeper teacher layers. Experiments on six video reasoning benchmarks show that Latent-OPD consistently outperforms output-only OPD. Notably, the improvements are particularly pronounced in scenarios with limited frames, long videos, or tasks requiring complex evidence aggregation. These results establish Latent-OPD as a highly effective approach to frame-efficient video reasoning.

Shared autonomy requires principled mechanisms for allocating and transferring control between a human and an autonomous agent. Existing approaches often rely on blending control inputs or heuristic switching rules, which lack theoretical guarantees and fail to account for the dynamics of authority transfer. This paper develops a cooperative game-theoretic framework for authority switching in shared autonomy. We formulate the control switching problem as an identical-interest dynamic game in which authority transitions are embedded into the system dynamics, yielding optimal switching policies rather than ad hoc rules. We establish the existence and characterization of team-optimal policies in pure strategies under stochastic human override, accounting for asymmetric authority where humans retain override capability. For linear-quadratic systems, we derive closed-form recursions for the optimal switching policies and value functions, enabling efficient computation independent of the continuous state. We validate the framework on scalar and multi-dimensional linear systems, demonstrating how optimal switching adapts to varying system dynamics, cost structures, and override probabilities. The results reveal fundamental trade-offs between human adaptability and autonomous efficiency, illustrating the practical benefits of grounding shared autonomy in cooperative game theory.

Long-term emotional-support agents require memory mechanisms for personalized understanding across sessions. However, emotional-support dialogue is often low-density: turns are incomplete, evidence is scattered, and user states evolve over time. Existing memory methods usually rely on fixed units, such as turn-level notes or session summaries, which may lose details or introduce redundant noise. We propose FTA-Mem, a structured memory framework for low-density long-term dialogue. FTA-Mem uses Boundary-preserving Window Segmentation (BWS) to form coherent situation fragments, and constructs Fact-Time-Affect Memory Units (FTA Units) that jointly encode factual content, temporal grounding, and affective context. Retrieved units are then synthesized into structured context for answer generation. Experiments on ES-MemEval and LoCoMo show that FTA-Mem improves overall long-term memory question answering across benchmarks with different information-density characteristics. On ES-MemEval, FTA-Mem achieves 0.3871 F1 and 0.6668 BERTScore. Further analysis shows that situation-level FTA construction better balances evidence preservation and construction cost than coarse session-level or overly fine-grained turn-pair construction, providing an effective granularity trade-off for long-term dialogue memory.

Shared autonomy requires principled mechanisms for allocating and transferring control between a human and an autonomous agent. Existing approaches often rely on blending control inputs or heuristic switching rules, which lack theoretical guarantees and fail to account for the dynamics of authority transfer. This paper develops a cooperative game-theoretic framework for authority switching in shared autonomy. We formulate the control switching problem as an identical-interest dynamic game in which authority transitions are embedded into the system dynamics, yielding optimal switching policies rather than ad hoc rules. We establish the existence and characterization of team-optimal policies in pure strategies under stochastic human override, accounting for asymmetric authority where humans retain override capability. For linear-quadratic systems, we derive closed-form recursions for the optimal switching policies and value functions, enabling efficient computation independent of the continuous state. We validate the framework on scalar and multi-dimensional linear systems, demonstrating how optimal switching adapts to varying system dynamics, cost structures, and override probabilities. The results reveal fundamental trade-offs between human adaptability and autonomous efficiency, illustrating the practical benefits of grounding shared autonomy in cooperative game theory.

While various organizations now actively encourage LLM use in classrooms, we still lack rigorous, systematic evaluations of how well these models actually perform the fundamental tasks of language pedagogy. This paper examines whether state-of-the-art LLMs can deliver the kind of corrective feedback and methodological explanations that language learners need. The study tests multiple large language models on their ability to identify, correct, and explain common learner mistakes in English, by systematically varying model parameters to investigate how these technical adjustments affect output quality, pedagogical clarity, and consistency, along with using retrieval-augmented generation to query methodological data. The evaluation employs automated metrics (GLEU, BERTScore) but also human expert judgments to capture dimensions that purely computational measures miss: linguistic nuance, cultural sensitivity, and instructional appropriateness. While models demonstrate impressive surface-level correction abilities, their explanations often lack the terminological and domain knowledge that effective language teaching requires, suggesting that current enthusiasm for AI-assisted language learning may be outpacing our understanding of these systems' actual pedagogical competence.

Differentiable nonlinear model predictive control (NMPC) provides a principled way to embed optimal control structure into end-to-end learning paradigms, but its practical use is often limited by the computational and memory costs of both forward optimization and backward sensitivity propagation. This brief proposes PANDA, a matrix-free solver for differentiable NMPC. In the forward pass, PANDA combines proximal-gradient iterations with quasi-Newton acceleration and introduces an adaptive stepsize enlargement mechanism to mitigate the conservativeness of monotone stepsize reduction. The resulting stepsize behavior and its effect on local convergence are theoretically analyzed. In the backward pass, PANDA performs implicit differentiation from the residual equation and computes adjoint sensitivities using Krylov-subspace iterative methods together with automatic-differentiation-based Matrix-Vector product operators, thereby avoiding explicit Hessian and Jacobian construction. The method is evaluated on a nonconvex trailer NMPC problem embedded in an imitation learning task. The results show that PANDA achieves much faster forward and backward computation and lower memory overhead than representative differentiable optimization solvers, while maintaining effective imitation learning performance.

Surface acoustic wave (SAW) micromanipulation enables the precise, non-contact handling of microscale particles and has attracted considerable interest in microfluidics and biomedicine. However, conventional SAW platforms generally rely on simple interference fields which are susceptible to fabrication imperfections and environmental perturbations, resulting in limited trapping stability. Here, we develop a SAW-based acoustofluidic platform that generates an acoustic skyrmion lattice through the coherent interference of three SAWs. The topologically structured field provides robust phase singularities and a stable gradient-force landscape, enabling microparticles to be localized at predefined lattice sites and supporting controllable rotational manipulation. Independent modulation of the amplitude and phase of the electrical inputs allows the field strength to be tuned for particles of different sizes. Numerical simulations and proof-of-concept experiments confirm particle trapping and ordered lattice assembly in the acoustic skyrmion field, demonstrating the feasibility of translating topological acoustic textures into practical on-chip manipulation functions. This reconfigurable strategy offers a route to robust SAW manipulation and may support applications in single-cell analysis, three-dimensional cell assembly, high-throughput screening, and microscale and nanoscale device assembly.

Coordinated multi-agent exploration requires not only efficient individual coverage but also non-redundant coverage across agents over extended planning horizons. Conventional approaches rely on hand-crafted coordination rules, while end-to-end multi-agent learning methods are difficult to scale and train. Diffusion-based planners such as DARE offer a promising alternative by generating long-horizon trajectories instead of single-step actions, but existing methods are trained on a narrow planner distribution, limiting behavioral diversity and inference-time controllability. We propose a Planner-Conditioned Diffusion Policy (PCDP) for graph-based multi-agent exploration. PCDP is trained on demonstrations from multiple planner styles with planner identity as an explicit conditioning input, enabling a single shared model to learn a multimodal trajectory distribution and generate diverse, controllable trajectory candidates from the same observation. Rather than learning coordination end-to-end, we reuse this multimodal single-agent policy across all agents and introduce coordination through local reranking, in which nearby agents jointly select the trajectory combination with minimal predicted overlap. We evaluate PCDP against classical and diffusion-based baselines on 100 held-out maps in a four-agent simulation setting. PCDP matches the perfect success rate of the diffusion-based baselines while improving mean max-agent travel, total team travel, and agent imbalance. Crucially, reranking alone over a single-planner baseline yields only marginal gains, indicating that planner-conditioned multimodality is the main contributor to improved coordination. Qualitative simulation results and real-robot experiments with two agents further validate that diverse long-horizon trajectory generation produces emergent spatial separation between agents without any explicit repulsion mechanism.

Humanoid intelligence requires learning over an extremely diverse space of whole-body motions and physically grounded interactions. However, existing embodied datasets remain fundamentally limited: internet-scale video data lack precise physical states and interaction grounding, while laboratory motion datasets provide high fidelity but only narrow behavioral coverage. This mismatch creates a critical bottleneck for scalable humanoid policy learning. We present HiPHI, a 600+ hour scale high-fidelity whole-body human motion dataset designed to systematically maximize coverage of the human motion and interaction manifold. HiPHI is theoretically guided by FrameNet, a linguistic framework organizing human primitives. Created using an optical motion capture pipeline, HiPHI provides sub-millimeter spatial marker tracking accuracy for full-body human motion and mesh-level object trajectories. We further introduce a benchmark suite evaluating motion-space diversity, interaction grounding, object consistency, and physical AI applications. Our analyses demonstrate that HiPHI significantly expands motion coverage compared to existing motion datasets while maintaining high-fidelity interaction quality, and establishes a scalable data foundation for training, evaluating, and generalizing humanoid policies in real-world embodied tasks, where similar extensions are also applicable to motion prior models in computer graphics.

In modern intelligent transportation systems, it is essential to accurately estimate vehicle positions, especially in mixed traffic conditions where both connected and conventional vehicles coexist. Roadside infrastructure and connected vehicles can provide complementary observations of the same traffic scene, but real-world evidence on decision-level fusion between these sources remains limited. This paper proposes a multi-observer vehicle localization framework that fuses compact object-level detections from a static roadside radar and a dynamic LiDAR-equipped connected vehicle. We evaluate the framework with real-world data collected at an urban intersection in Helsinki, Finland, with a separately instrumented target vehicle used as the reference trajectory. Two extended Kalman filter based strategies for the localization task were benchmarked. The performance of the radar and LiDAR sensors were evaluated separately, and the two fusion strategies were explored under nominal sensing conditions, reduced LiDAR update rates, simulated LiDAR occlusions, and different target-vehicle motion states. The results show that, under full LiDAR availability, fusion performance is dominated by the LiDAR observations, while the less accurate and less consistent radar observations provide only limited additional improvement. Nevertheless, AEKF achieves small gains over the LiDAR-only baseline, and object-level connected vehicle observations remain useful when shared at reduced update rates. These findings indicate that decision-level fusion provides scenario-dependent benefits rather than automatic improvement over a strong single-sensor baseline. We release the dataset and implementation on Github to support further research: https://github.com/AppuriAalto/multi-observer-vehicle-tracking

Achieving human-level competitive intelligence and physical agility in humanoid robots remains a profound challenge, particularly in contact-rich and highly dynamic tasks such as boxing. While Multi-Agent Reinforcement Learning offers a principled framework for strategic interaction, its direct application to unstructured raw motor spaces inevitably leads to joint-level physical collapse, preventing the emergence of any viable combat tactics. To resolve this fundamental conflict between strategic exploration and physical feasibility, we formulate the humanoid combat task as a novel two-player latent-space zero-sum Markov game. Under standard regularity and approximate best-response assumptions, we show that the latent formulation induces an equivalent game over the decoder-reachable action manifold, providing an approximate-Nash interpretation of the resulting self-play dynamics. To instantiate this theoretical formulation, we propose RoboStriker, a hierarchical framework that decouples high-level reasoning from low-level execution. It first distills the tracking expertise of predefined boxing motions into a topologically bounded latent manifold. This structured latent foundation subsequently drives multi-agent co-evolution via Latent-Space Neural Fictitious Self-Play. Extensive experimental results demonstrate that gaming within this structured latent space substantially outperforms direct exploration. By constraining strategic exploration through a pretrained motion decoder, RoboStriker substantially reduces the catastrophic balance failures observed in raw action-space methods and achieves superior tactical performance in both competitive win rates and striking efficiency. Finally, we successfully deploy and validate our learned combat policies on real-world humanoid robots. Our code and video and supplementary materials are available at RoboStriker.

At every re-observation point in a hierarchical Vision-Language-Action (VLA) system, two interface decisions must be made: when to terminate the current subtask and how far to execute the proposed action chunk. These decisions are mutually dependent---the optimal stopping point depends on what the executor plans to do, while the optimal execution length depends on where the subtask boundary lies---yet existing architectures evaluate them in isolation, an asymmetry neither module can overcome alone. We present SparkVLA, a stop-aware hierarchical VLA that resolves this mutual dependency by formulating both decisions as a single ranking: Stop competes against every action-prefix length in a unified candidate set, and the system selects the highest-scoring option, eliminating threshold tuning and requiring only offline ordinal preferences. An Anchor-Conditioned Context Encoding module caches a history-aware subtask anchor encoding onset-state memory and goal semantics, guiding visual-token pruning toward task-relevant regions; a Stop-Aware Action-Prefix Selection head scores all candidates via full self bnattention at chunk boundaries for efficiency. On RoboCerebra, SparkVLA achieves 47.12% success rate, surpassing the official hierarchical baseline by 30.57% and the strongest reproducible method by 26.83% Real-robot experiments on multi-step tasks further validate these gains on physical hardware.

Training locomotion policies for complex unstructured terrain requires a curriculum to avoid early exploration failures. However, since unstructured terrain lacks explicit difficulty ordering for curriculum design, existing methods resort to heuristic curricula over parameterized terrains. This abstraction limits generalization, as policies can overadapt to near-fixed perceptual patterns. To address this, we propose \textbf{\ourname{}}, an \textbf{T}rajectory-level \textbf{A}utomatic \textbf{C}urriculum \textbf{L}earning framework that generates training tasks directly from unstructured terrain maps. At each curriculum update, the evaluator learns a difficulty function for the current policy that maps a given trajectory task to a difficulty score. The sampler then proposes new trajectories guided by the learned evaluator as the curriculum for the next policy update. This forms a closed loop in which the curriculum is iteratively matched to the evolving policy. Quantitative and qualitative experiments show that \ourname{} continuously provides effective curricula on unstructured terrain, improving trajectory success rate by \(56.3\%\) over direct training without curriculum. Compared with handcrafted curriculum learning, our method improves success rate by \(18.5\%\) on the hardest terrain tasks and by up to \(39.74\%\) when evaluating traversal from diverse approach directions on the same obstacle type.

Robot manipulation requires policies that are both accurate and efficient, as robot control must respond to changing observations under tight latency constraints. Recent diffusion and flow policies are promising, but they often treat conditions as auxiliary signals rather than jointly evolving them with action trajectories. We find that this limitation can be effectively mitigated by a \textbf{simple yet effective unified condition-action modeling design} that represents conditions and actions in a shared token space, allowing a compact model to achieve high performance while improving both inference speed and accuracy. Therefore, we propose UCA-Flow, a unified condition-action modeling framework for accurate one-step action generation. Our method unifies observation conditions, timestep conditions, interval conditions, and action tokens into a single sequence, and processes them with a Unified Condition-Action Transformer for joint condition-action representation learning. As a result, condition representations are dynamically reconstructed according to the current generation stage, highlighting information most relevant for action refinement. Furthermore, we introduce an improved dual-pass supervision scheme over $u$ and $v$ for stronger optimization of unified condition-action modeling. UCA-Flow improves the average success rate by 9.3 percentage points over the strongest baseline, while achieving $45.6\times$ and $33.4\times$ speedups over DP3 and Simple DP3, and remaining $4.3\times$ and $2.3\times$ faster than one-step FlowPolicy and MP1, respectively.

Valley photonic crystals enable valley-dependent transport and chirality-selective emission, but incorporating wavelength-scale localization remains challenging. Existing valley-photonic-crystal cavities rely on finite defects or local lattice modifications that require structure-specific optimization and offer limited continuous control. Here, we theoretically and experimentally demonstrate two-dimensional nanocavity confinement using two orthogonal domain walls in a glide-symmetric valley photonic crystal. A valley domain wall confines the guided interface mode transversely, while an SSH-like domain wall localizes it longitudinally. Starting from a glide-symmetry-protected Dirac point in a bearded-interface waveguide, controlled displacements of adjacent triangular holes open a topological gap in the continuous guided-mode dispersion. The displacement amplitude $ΔR$ tunes the gap, mode volume, and intrinsic radiative $Q$ factor. Implemented in a silicon photonic-crystal slab, the structure exhibits localized resonances within the topological mode gap and systematic spectral tuning with $ΔR$. The maximum measured loaded $Q$ factor is $1.2\times10^{4}$. This approach enables continuously tunable, high-$Q$ nanocavities integrated into topological waveguide networks for compact resonant devices and enhanced light--matter interactions.

Electric vehicles (EVs) represent a critical intelligent terminal within the Internet of Things (IoT). Despite the year-on-year growth in EV penetration, the highway driving experience still requires improvement. Accurate prediction of EV highway charging behavior is crucial to addressing this issue. This paper introduces a novel bounded rationality framework to analyze highway charging decisions. Specifically, we utilize prospect theory to capture the tendency of drivers to reserve more electricity than theoretically necessary. We then propose a Bayesian game in which EV drivers, unaware of others' decisions, aim to minimize costs, including range anxiety, charging fees, and queuing time. To gain insights into the game, we prove the existence and uniqueness of the Bayesian Nash Equilibrium in two practical scenarios. Our numerical experiments, based on real-life data, demonstrate that drivers' risk aversion tendency significantly influence EV charging decisions, charging demand, queuing lengths at charging stations, and the departure rate on the highway network. Furthermore, our strategy reduces cumulative EV cost and CSs' charging costs compared to other benchmarks.

Artificial intelligence-assisted ultrasound scanning enhances diagnostic reliability and efficiency by providing real-time guidance for standardized image acquisition and reducing operator dependence. However, existing reinforcement learning and learning-assisted ultrasound scanning methods typically rely on carefully designed reward functions or extensive interaction data, which limits their generalization ability and stability across different devices, patient populations, and complex clinical scenarios. To address these challenges, we propose an ultrasound vision-language-action model (US-VLA) for automated ultrasound scanning that explicitly encodes clinical semantic goals and generates sequential probe manipulation actions under real-time ultrasound feedback. In particular, we first design an ultrasound-aware expert fusion module to jointly integrate ultrasound observations with auxiliary contextual information, enabling semantic ultrasound feedback to effectively guide the scanning process. Then, we construct US-VLA-Data, a real-world dataset covering liver and kidney examinations, which includes five clinically defined standard planes and comprises 320 expert scanning trajectories with approximately 80,000 synchronized timesteps. Extensive experiments demonstrate that US-VLA achieves competitive performance in ultrasound probe manipulation tasks, indicating its effectiveness and promising generalization within the evaluated abdominal ultrasound setting. The source code is available at https://github.com/VMVLab/US-VLA.

Learning-based surgical robot autonomy requires large-scale demonstrations with synchronized videos and robot actions, but such data are exceedingly rare in clinical or realistic tissue settings because robot kinematics are typically inaccessible outside controlled research systems. In contrast, phantom data collected on research platforms provide accurate action labels but lack the visual diversity of real tissue. We propose SurgVIL, a framework for scaling surgical robot imitation learning using open-source surgical videos. SurgVIL combines kinematically labeled phantom robot demonstrations with surgical videos from open-source datasets and online sources for policy learning. Since these videos lack robot motion labels, we estimate approximate kinematics as weak supervision. We evaluate SurgVIL on two da Vinci robot tasks: needle pick-up and cholecystectomy cutting. Across ACT, $π_0$, and GR00T-H backbones, adding surgical videos substantially improves generalization to real-tissue and out-of-distribution settings, suggesting a scalable path from phantom training toward generalizable surgical robot policies.

Semiconductor quantum-device design still lacks an integrated Technology Computer-Aided Design (TCAD)-like environment that connects material geometry, quantum many-body simulation, and automated design. Here we introduce zenDot, a large-language model (LLM)-integrated quantum TCAD platform that links a material-labelled device state to a unified condensed-matter physics toolbox. The device and calculation components are integrated into a desktop workbench, Python API, and an embedded LLM agent, allowing electrostatics, charge and transport characterization, correlated-state calculations, and qubit modelling to be executed within one reproducible environment. We demonstrate zenDot on a Si/SiO2 double quantum dot, where a single device state reproduces the characterization workflow and supports hybrid, tunnel-charge, and singlet-triplet qubit analyses. A platform-level universal-control scan revises the singlet-triplet operating point and reduces the predicted worst-gate infidelity by nearly 30-fold. Beyond analysis, the LLM agent directly operates the same physics environment as human users, proposing design changes, executing registered simulations, and iterating on solver-returned metrics under physics-aware validation. Across three demonstration tasks it completes 18 validated design iterations, including geometry modification followed by a full re-solve from the material stack. zenDot establishes a machine-operable quantum TCAD workflow that connects device physics with LLM-driven design exploration.

We introduce ScenarioCharacterization, an open-source framework for automated, dataset-agnostic profiling of driving scenarios in trajectory datasets. Our framework is packaged as a modular, configuration-driven pipeline of three layers: a dataset adapter that maps custom datasets onto an open Scenario representation, a characterizer that performs feature extraction, behavior probing, and criticality scoring at scenario and agent levels, and an analysis layer for scenario visualization and feature, score, and probe analyses. Because the layers communicate only through Pydantic-validated schemas composed via configurations, a new dataset can easily plug in without rewriting the characterization and analysis stack. This technical report describes the design and APIs, shows example outputs on Waymo Open Motion, Argoverse2, and nuPlan, and discusses downstream uses of the approach. The framework is available at https://github.com/navarrs/ScenarioCharacterization.

Large language models (LLMs) are increasingly used to generate textual robot design specifications, interaction policies, and risk assessments during early-stage robot development. Such outputs may influence how surveillance and security robots are conceptualized, documented, and ultimately implemented. This paper evaluates whether identity-conditioned prompts produce systematic differences in LLM-generated surveillance and security robot design descriptions. Using 236 demographic identity labels across single-label and model-augmented prompt conditions, we analyze readability as an initial benchmark for evaluating accessibility and identity-conditioned variation in generated robot design descriptions. The results show significant differences in readability across prompt conditions, design dimensions, and demographic identities. Although readability cannot determine whether an output is fair or socially appropriate, it provides an interpretable baseline within a broader benchmarking framework that also includes lexical, semantic, sentiment, syntactic, and fairness-focused analyses.

There has been a flurry of activity in the development of robotic systems to localize and track underwater man-made or natural targets based on sparse acoustic data. Compelling examples include the development of surface tracking systems to aid in the navigation of groups of underwater vehicles performing environmental monitoring missions or to study the motion patterns of large underwater fauna. With current technology, the latter case can only be tackled using Time-Difference-of-Arrival (TDoA) techniques. Recent progress in nonlinear state estimation indicates that optimization-based methods may overcome the limitations of classical recursive filtering. However, achieving reliable estimator performance in the case of nonlinear target dynamics and sparse measurements remains a key challenge. In this paper, we study a Moving Horizon Estimation (MHE) approach to TDoA-based underwater target tracking. Through a 2D simulation environment capturing typical marine conditions, we show that the MHE-based estimator maintains reliable tracking in the considered scenarios even when the classical EKF becomes unreliable. The results highlight that multi-step trajectory coupling and physically consistent constraints, which are key advantages of the MHE approach, significantly enhance estimator robustness. It is shown that the MHE approach offers promise as a practical and scalable building block for future multi-agent tracking systems based on TDoA measurements operating in real underwater missions.

Modern power systems and smart grids are often composed of fragmented and heterogeneous data silos, which lack the cohesion needed for effective cross-domain analysis. For this, this paper introduces a universal ontology framework for the operational representation of intelligent cyber-physical power systems via a unified knowledge graph and an ontology capable of cross-domain reasoning. This work focuses on bridging cyber-physical simulators as a stepping stone towards that vision. By establishing a unified semantic middleware grounded in IEC 61970 (CIM) and IEC 62351/61850 standards, this framework integrates disparate cyber and physical simulation environments, illustrated via OMNeT++ and PowerWorld, into a single knowledge graph. Evaluation across three standard power system benchmarks demonstrates sub-linear scaling in both knowledge graph size and construction time. We further validate the framework's efficacy for real-time decision support, achieving millisecond-level query performance across both domains, maintained across six cumulative structural mutations to the knowledge graph. The resulting unified knowledge graph provides a robust, scalable information corpus for autonomous smart grid operations, enabling complex analysis of real-world power systems.

Upper extremity motor function recovery is positively linked to Task-Specific Training (TST) and sufficient therapy dosage. Rehabilitation robots can increase TST dosage via controlled, repetitive treatment and free therapists to simultaneously manage other patients, but it has yet to demonstrate significant benefits over conventional treatment. This is potentially linked to inaccurate robotic representation of personalised physical therapist-patient interaction and lack of practice variability during TST. Hence, we advocate for robotic interventions that preserve the personalised physical therapist-patient interactions when delivering TST for patients across varying practise conditions. We propose a Learning-from-Demonstration framework using Task-Parameterised Gaussian Mixture Models (TPGMM) to learn personalised physical therapist-patient interaction in Task-Specific exercises, mapping patient joint kinematics to therapist-applied torques using few demonstrations. The model is generalised to reconstruct therapist torques in new task variations. The framework was evaluated on physical interactions from 14 mock "therapist-patient" pairs over three tasks of increasing complexity, each with six variations. A benchmark comparison against a Look-Up Table was conducted. The results show both methods reproducing interactions in unseen task variations that deviate slightly from the actual interaction, with TPGMM slightly outperforming LUT. Both methods reproduced interactions that gets increasingly closer to the actual interaction as task complexity increases.

Low-cost four-degree-of-freedom (DoF) arms are among the most accessible robotic platforms. But they are, in theory, underactuated for picking up in situations where objects are at arbitrary orientations, a task that appears to require five degrees of freedom: the planar position (x and y), the height (z), a wrist rotation to align the gripper with the object, and gripper actuation, of which a four-DoF arm lacks the wrist rotation. This work shows that perception and motion planning can enable such an arm, a roughly $200 Waveshare RoArm-M2-S, under a fixed overhead camera to detect and color-sort writing utensils without that joint. A YOLO11n-OBB (You Only Look Once, oriented bounding box) detector locates each writing utensil; camera intrinsics and an ArUco reference pose convert its pixel coordinates to robot coordinates; and a color classifier labels it. The detected orientation angle determines the motion strategy: utensils close to the arm's fixed approach direction are picked up directly, and those at steeper angles are reoriented via corrective sweeps until they are graspable, after which they are picked up and sorted into the assigned color bin. Across 326 logged motions on seven writing utensils, the arm made 196 direct grasps and 130 corrective sweep passes, correcting misalignments up to 90 degrees, suggesting that clever task-informed engineering can compensate for a missing degree of freedom on tasks like this one.

Rerouting the tendon path in tendon driven continuum manipulators (TDCMs) enables a broad range of deformation modes. This work presents a Reconfigurable TDCM design which allows independent rotation of intermediate spacer disks, thereby locally rerouting the tendon and achieving non-trivial backbone spatial deformations. Two such designs, (a) Manual Disk Locked (MDL) and (b) Continuous Disk Rotor (CDR) manipulators are presented to achieve disk rotations before and during operation, respectively. A predictive static model based on the piecewise constant strain (PCS) assumption is developed within a potential energy minimization framework, incorporating (a) disk rotations, (b) discrete tendon paths between disk segments, (c) rigid thickness of spacer disks, and (d) elasticity of the tendons. The model is validated against experimental results, demonstrating an average tip error of $1.2\%$ of the manipulator's total length for parallel tendon routing and around $3\%$ for the case when multiple disks are rotated. The computation time is an order of magnitude lower than the state of the art Cosserat rod solver.

Action chunking --- the practice of predicting a sequence of actions and executing a prefix open-loop --- has emerged as a key enabler of recent progress in imitation learning for robotic manipulation. However, executing long open-loop prefixes reduces reactivity, limiting policies' ability to correct for errors. Further, the mechanisms underlying these performance benefits remain poorly understood: prior works cite mitigating compounding errors, absorbing inference latency, or smoothing motions, but provide limited controlled evidence or guidance for preserving reactivity. In this work, we argue that long open-loop execution primarily helps short-context policies imitate "non-Markovian demonstrations". Across four simulation and two real-world tasks, we show that expert non-Markovianity strongly shapes the relationship between task success and open-loop execution horizon. Further, we investigate the impact of compounding errors --- the prevailing explanation for long open-loop execution in prior work --- and find that while they matter, expert non-Markovianity has a much stronger impact in our experimental setting. Finally, we show that when policies are provided with a sufficiently long context, open-loop execution is no longer beneficial and the most reactive, closed-loop policies perform best. While imitation learning has seen great success using long open-loop execution, our findings motivate long-context, reactive policies as a more principled and performant paradigm.

Functional electro-optic (EO) tunable photonic integrated circuits are crucial to next-generation information processing and advanced computing, where ferroelectric materials such as lithium niobate (LiNbO3) provide outstanding optical properties and versatile tuning mechanisms. However, beyond achieving high-performance devices, practical deployment of LiNbO3 photonic integrated circuits requires rapid and cost-effective scale-up, which remains challenging because of the fabrication complexity and high costs of ion-sliced thin-film lithium niobate and wafer bonding processes. To address this gap, we demonstrate a fully monolithic photonic platform, amorphous silicon carbide (a-SiC) on bulk LiNbO3 crystal substrate, enabling scalable, CMOS-compatible, low-cost, and high-density photonic integrated circuits without relying on thin-film LiNbO3. This integration approach breaks the long-standing cost-performance-scalability tradeoff, and intrinsically eliminates the need for sophisticated fabrication processes, including ion slicing, wafer bonding, and LiNbO3 etching. It realizes high-density, low-loss, and record high EO tuning efficiency of VpiL = 2.87 Vcm on bulk LiNbO3 crystal substrate. Furthermore, leveraging slow-light effect near photonic crystal bandgap, an 8.5-fold EO tuning efficiency enhancement is achieved, highlighting the platform's capability in confinement control and dispersion engineering.

Large-scale pre-training has made robot policy fine-tuning increasingly data-efficient, but this progress has largely been driven by datasets and embodiments built around simple parallel-jaw grippers. Dexterous, multi-fingered hands remain comparatively data-starved because real teleoperation is costly to scale, while human hand video is off-embodiment and requires lossy pose estimation and retargeting. We introduce Simulation Pre-training for Dexterity (SPD), a pre-training framework for dexterous manipulation that uses data entirely collected in simulation. In SPD, humans manipulate virtual objects inside a VR headset, enabling on-embodiment trajectories and robot-free collection. With the help of five operators, we collect 75 hours of multi-task dexterous manipulation over one week, and use it to pre-train a causal transformer on a sequence modeling objective. We study the benefits of simulation pre-training on real-world tasks by fine-tuning on 1-2 hours of physical demonstrations on a 56-DoF bimanual dexterous setup. We find that our approach outperforms training behavior cloning policies from scratch, showing that simulation teleoperation is a viable pre-training source for real-world dexterous manipulation. We perform ablation studies, measuring the benefits of history conditioning and short action chunks for reactive control.

Robot sensor designs, particularly tactile sensors, are highly diverse and evolve rapidly. Modeling each sensor in simulation demands substantial domain expertise and computational approximations can degrade the fidelity of the simulated signals. We propose Sim2Real via Bottlenecked Latent Reconstruction (SBLR), a framework that avoids sensor-specific simulation entirely by (1) training policies on a simulator-native oracle sensor that is easy to construct without modeling any particular sensor (e.g. we use a point-cloud and finger-tip forces as a tactile oracle), and (2) aligning real sensor latent embeddings to those of the oracle sensor at inference time. Policy training proceeds in two-stage: the policy first learns from the oracle sensor latents, then a bottlenecked latent reconstruction adapts it to the information loss expected when using the real sensor instead of the oracle. The alignment between oracle and real sensor is learned from unpaired random-play data collected in both simulation and the real world, using rectified-flow-based transformation networks trained on nearest-neighbor pseudo-pairs. Simulation experiments on three contact-rich tasks show that SBLR matches or approaches the performance of an oracle with direct access to tactile simulation. Hardware experiments on Peg Insertion and Gear Meshing with GelSight Mini and DIGIT sensors demonstrate 85-97.5% zero-shot success without requiring any sensor-specific modeling or calibration, outperforming a physics-based tactile simulation baseline by 7.5-15%.

Vision-language-action (VLA) models have become a dominant paradigm for generalist embodied agents, demonstrating strong complex and long-horizon task completion in structured settings. Yet it remains an open question whether current VLA systems can benefit from more effective architectural design, scale to substantially larger and more heterogeneous data regimes, and achieve broader generalization across tasks and embodiments. To this end, we present GigaBrain-0.7, an embodied foundation model with substantially improved generalization across diverse robot embodiments. Specifically, GigaBrain-0.7 unifies understanding, prediction, and action through a three-system architecture, scales pretraining to over 37,000 hours of heterogeneous embodied data, and introduces one-stage alignment training that jointly optimizes vision-language understanding and multi-embodiment action generation. Compared with the preceding GigaBrain-0 series and prior state-of-the-art models including $π_{0.5}$, GigaBrain-0.7 achieves substantial improvements in foundation zero-shot capabilities, language-conditioned instruction following, and post-training task success rates. In particular, on our in-house Maker H01 platform and mainstream robot embodiments, GigaBrain-0.7 demonstrates strong task adaptability and completion ability across both home and industrial scenarios. All training code and pretrained model weights will be released.

Operating household appliances requires long-horizon planning that is state-dependent and robust to disturbances, yet existing large models fall short, as no sufficiently diverse, task-oriented dataset exists to support such planning. To bridge this gap, we propose MAGE, a scalable data synthesis pipeline that introduces a novel Hierarchical Appliance Graph (HAG) to automatically generate part grounding, long-horizon planning, and closed-loop recovery data from appliance manuals. With MAGE, we build UseAppliance, the first large-scale dataset for manual-grounded appliance manipulation planning, spanning 22 appliance categories with 89K+ part annotations, 53K+ manipulation tasks, and 33K+ closed-loop adjustment steps. Built on UseAppliance, we develop AppliancePlan, an end-to-end model for manual-grounded appliance manipulation planning. On RealAppliance-Bench, AppliancePlan with only 7B parameters achieves over 10x the best baseline on open-loop planning and consistently outperforms state-of-the-art models across all tasks. Real-robot experiments on six household appliances further confirm effective sim-to-real transfer, marking an important step toward general-purpose household robotics.

As Vision-Language-Action (VLA) models scale toward real-world deployment, contact-rich manipulation exposes a critical blind spot: these policies encode broad visual-semantic priors yet remain unaware of local contact events, producing identical actions whether contact is established, lost, or destabilized. Existing remedies either modify VLA internals, risking catastrophic forgetting, or demand online reinforcement under near-failure contact conditions. Both grant tactile unbounded influence over action generation, conflicting with the priors that make VLAs generalizable. We introduce ViTaR, which reframes tactile feedback from an action-generating perceptual input to an execution modulator that selects and scales bounded residual corrections atop a frozen VLA, preserving pretrained capabilities by construction. ViTaR decomposes adaptation into two stages: Effect-Guided Modeling determines whether and which correction is locally justified via outcome-grounded preference evidence, and Residual Action Modulation converts this evidence into a residual choice with continuously scaled gain from real-time visuotactile observations. On the UniVTAC benchmark spanning seven contact-rich tasks, ViTaR achieves 61.3% average success, a 30.6 percentage-point improvement over its frozen VLA base that also surpasses purpose-built tactile baselines. Physical-robot experiments confirm that bounded tactile modulation transfers to real sensor noise and dynamics.

Humanoid robots are expected to traverse complex terrains, where the plantar support may vary dramatically due to foot placement errors, ground properties, and transient dynamics. To achieve robust locomotion, the robots are required to adapt to uneven terrain and uncertain foot--ground interactions. Existing locomotion policies rely primarily on proprioception or exteroceptive terrain perception, where the former provides only indirect evidence of plantar support, while the latter predicts contact conditions before touchdown but cannot observe the actual support in real-time. Although some studies incorporate plantar contacts as an auxiliary perception, they rely mainly on summary statistics, overlooking the spatial topology of plantar pressure, which provides a more direct characterization of the realized contact state. To bridge this gap, we present Tac4Loco, a tactile-perceptive framework that incorporates multi-array plantar pressure as direct feedback for humanoid locomotion. We formulate a topology-preserving ordinal representation to map simulated and physical sensor signals into a shared observation space, with a dual-branch encoder for extracting their spatial and temporal representations. Subsequently, the learned spatiotemporal features are integrated with augmented proprioception including terrain estimation cues, and provided to an asymmetric actor-critic architecture for policy learning. Extensive simulation and real-world experiments demonstrate improved tracking performance and support adaptation on terrains with inclined, partial, asymmetric, and changing support. We further demonstrate its zero-shot deployment on unseen compliant and unstructured terrains, including a foam platform and a gravel road. All code and experimental configurations will be released as open-source to facilitate reproducibility.

A dual-representation flow-matching policy decodes each predicted motion into joint and end-effector spaces, and the residual between the two kinematically equivalent decodings provides a physically interpretable runtime signal. On multimodal tasks, however, independently sampled branches may choose different valid modes, causing false alarms. We study how to coordinate the two branches and at what cost. Across two robot environments and a non-robotic testbed, the tested mechanisms fall into four classes. An auxiliary latent shared by both branches but absent from the flow-matching construction is erased at the population optimum, a provable dead end confirmed within a prespecified 2% equivalence band. Sharing source noise can coordinate or anti-coordinate: its effect changes sign with the representation map and tracks the alignment of decoder mode basins. Consistency regularization gives intermediate coordination but reduces the valid-pair rate, while training-supported discrete partitions achieve near-ceiling coordination robustly. We further derive a chance-corrected coordination bound based only on each branch's Gini-Simpson diversity, yielding an attainable region and a label-free certificate that separates coordination from collapse when zero mismatch is ambiguous. On LIBERO-Plus, benign multimodality adds 1.57 percentage points of false alarms to the residual, which remains the strongest evaluated failure signal; the preregistered token intervention does not meet its false-alarm criterion or produce a seed-robust detection change. Code, models, and per-run configurations are available at https://github.com/kimo423/dual-head-coordination.

Post-disaster damage to terrestrial infrastructure can disrupt wireless coverage,while Uncrewed Aerial Vehicle (UAV) swarms provide a promising solution for rapid restoration.However, due to the limitations in local geometry observations hidden radio impact,and inter-UAV communication,there exists a significant gap between locally visible movement choices and swarm-level coverage outcomes.To combat this gap,we propose a raido World-model-based Optimized Negotiation framework for Distributed UAV covERage (WONDER).Particularly, to tackle the unavailability of the future radio field from onboard observations, WONDER uses a Joint-Embedding Predictive Architecture (JEPA)-based radio world model to learn and predict the incremental radio effect of each candidate trajectory from deployment-available information.Multi-round negotiation in WONDER then coordinates ranked proposals by committing one trajectory at a time and re-evaluating the remaining proposals under the updated context. Our theoretical analyses further validate the effectiveness of such a world model-based framework. WONDER also adopts a Proximal Policy Optimization (PPO)-style Actor and alternates between updating the world model and the actor. Furthermore,we build RadioDynamics,a comprehensive simulation environment that integrates UAV mobility,radio propagation, inter-UAV communication modeling,and digital-twin geometry with ray-traced fields in $62$ metropolitan scenes.Experiments on $11$ testing scenes in RadioDynamics show that WONDER achieves the highest balanced score among seven evaluated methods,reaching $0.870$ with a $0.162$ coverage advantage over STACCA, while maintaining $100\%$ connectivity between UAVs.

What happens when you teach an LLM-based agent the scientific method? Motivation: Scientific discovery emerges from cycles of hypothesis, implementation, empirical testing, and feedback. Can this process be automated? We approach automated algorithm design through the lens of the scientific method, where an LLM-based agent goes through each step of the process in an ordered, iterative fashion. Results: We present The Little Scientist, a framework in which a "Scientist agent" works inside an evaluation environment that benchmarks its code and returns structured per-instance diagnostics. When the Scientist plateaus at a local optimum, a "Kuhn agent" injects a paradigm-shifting conjecture paired with a cross-disciplinary inspiration, forcing exploration of a different region of the LLM's latent space. We demonstrate the framework on two problems that require fundamentally different modes of discovery. For protein fitness prediction, the Scientist discovered Delta V, an ensemble calibration strategy that ranks first on the ProteinGym DMS Substitutions Zero-Shot leaderboard across all five official evaluation metrics, exceeding the #2 model (VenusREM) by +0.033 mean Spearman correlation across 217 DMS assays. For DNA motif discovery, the Scientist wrote an algorithm from scratch--DALE (Dual-seed Algorithm for Latent Enumeration)--that outperforms STREME (the default in the MEME Suite) across 132 ENCODE transcription factors (mean AUROC 0.842 vs. 0.803, Wilcoxon p < 10^{-6}) while running 11x faster. This demonstrates that the framework can produce genuinely novel algorithms, not just optimize existing components. Together, these results show that an LLM agent stepping through the scientific method can discover both new algorithms and new ensemble strategies that outperform prior solutions. The entire research program consumed 704M tokens on a single virtual machine with no GPUs

The rapid adoption of large language models has enabled the development of clinical multi-agent systems (MAS) capable of integrating multimodal patient data and supporting increasingly complex clinical decision-making. However, the deployment of these systems in real-world healthcare settings raises critical ethical concerns related to safety, fairness, accountability, transparency, and patient trust. While numerous organizations, including the World Health Organization, the National Academy of Medicine, and the FUTURE-AI consortium, have proposed ethical frameworks and governance principles for healthcare AI, these efforts remain largely conceptual. To address this challenge, we present ETHOS (Ethics and Trust through Hierarchical Oversight System), a modular ethics framework designed as a governance meta-agent that can be integrated with any existing multi-agent system without requiring changes to its underlying architecture. ETHOS translates stakeholder-informed ethical requirements into executable runtime oversight through a layered governance approach consisting of deterministic checks, contextual reviews, and a final ethics critic. These components continuously evaluate intermediate reasoning steps and final outputs, enabling the system to identify ethical risks, request revisions, or suppress responses that fail predefined safety and trustworthiness criteria. We demonstrate ETHOS within a hepatology clinical decision-support MAS. Results show that ETHOS improves decision reliability by detecting incomplete, inconsistent, or out-of-scope evidence and appropriately increasing abstention when safe recommendations cannot be supported. By embedding ethical governance directly into system operation, ETHOS provides a practical and auditable mechanism for transforming high-level AI ethics principles into deployable safeguards.

Reasoning segmentation enables vision-language models (VLMs) to translate mission-relevant language requests into pixel-level visual grounding, offering a natural perception interface for embodied agents. However, existing benchmarks largely focus on generic visual scenes and overlook the domain and resource constraints encountered in flood-response platforms. We present FloodReasonBench, a benchmark for VLM reasoning segmentation for embodied flood response at the edge. At its core, FloodReasonBench introduces FloodResponseSeg, a flood-specific reasoning-segmentation dataset constructed from real-world scenes and response-relevant targets. Beyond task accuracy, the benchmark characterizes reasoning-segmentation pipelines under lightweight visual encoding, hierarchical split inference, and compressed intermediate representations. We observe strong partition-dependent accuracy variation in the generic pre-adaptation setting, while the flood-adapted target-workload design space exhibits a substantially more compact accuracy range across partitions. Evaluation on an NVIDIA Jetson AGX Xavier further exposes the tradeoffs among reasoning-segmentation accuracy, edge-side latency, energy, and communication footprint, enabling quality-constrained selection of edge operating points. Together, these results provide a task- and system-level characterization of reasoning segmentation for resource-constrained embodied flood response at the edge.

Large Language Models (LLMs) are notorious for struggling with assessing their own uncertainty, detecting knowledge conflicts, or recognizing when problems exceed their expertise; such limitations inevitably undermine reliability and trust in LLMs. In this paper, we present the first implementation of a metacognitive framework for ensembles of LLMs that addresses these challenges through explicit monitoring and control mechanisms. Our system computes a Metacognitive State Vector (MSV) quantifying self-awareness for monitoring across five dimensions derived from cognitive psychology: Emotional Response, Correctness Evaluation, Experiential Match, Conflicting Information, and Problem Importance. MSV values also provide self-regulation for control, automatically switching between System 1 (fast, single- or multi-node) and System 2 (deliberative, multi-node) processing based on query complexity. For System 2 execution, graph-theoretic algorithms control the assignment of specialized roles (Domain Expert, Critic, Evaluator, Synthesizer, and Generalist) to ensemble nodes according to their MSV-quantified metacognitive states. Our implementation allows users to explore how different query types trigger distinct processing modes. The Proof-of-Concept (PoC) demo showcases the framework with illustrative examples showing appropriate System 1/System 2 routing and helps visualize the metacognitive process via real-time radar charts and decision indicators. This PoC implementation demonstrates the feasibility of creating a framework for metacognitive self-awareness and self-regulation in LLM systems.

We study generating game-theoretically optimized Courses of Action (COAs) for a Blue UAS swarm against an adaptive Red adversary in a communication-degraded environment, motivated by (but not derived from) a public U.S. Air Force SBIR solicitation. We propose UC-PSRO (Utility-Conditioned Policy-Space Response Oracles with a Communication-Dropout Curriculum), combining three mechanisms: (i) PSRO self-play, so Blue and Red policies train as approximate best responses to each other rather than one side against a fixed scripted opponent; (ii) FiLM conditioning of the Blue policy on a Commander's-Intent weight vector, sampled from a Dirichlet distribution during training, so one trained policy is re-steerable at execution time without retraining; and (iii) a curriculum annealing communication-graph edge dropout during training, so the swarm learns decentralized, peer-to-peer fallback instead of depending on full connectivity. We evaluate on a synthetic, unclassified stand-in for the solicitation's maritime scenario, with 5 seeds at N=25 Blue agents and a scalability sweep to N=200. We find a genuine trade-off, not a uniform win: the communication-dropout curriculum alone gives the strongest, most robust mission-completion rates of any learned method, improving counter-intuitively as denial increases (35% to 62% success as dropout rises from 0 to 0.75); adding utility-conditioning and PSRO self-play substantially slows convergence within a fixed budget, and we find no reliable exploitability advantage for self-play over a fixed-opponent policy, both statistically indistinguishable from a small, near-zero gap. We report this honestly as a convergence cost not yet offset by a demonstrated robustness benefit, rather than overstating one method as dominant, and provide a fully vectorized, open environment training at N=200 agents in single-digit milliseconds per step on a single consumer GPU.

Control barrier functions (CBFs) provide a systematic framework for enforcing safety constraints in nonlinear control systems. However, their implementation typically relies on accurate system models, which can limit their applicability in the presence of significant modeling uncertainties or unknown dynamics. This paper proposes a model-free framework for the computation of recursive control barrier functions based on the ultra-local model approach that leverages online estimation of the unknown system dynamics to construct CBF constraints. This approach does not require an explicit model of the system dynamics and enhances robustness with respect to disturbances and model mismatch. The resulting control architecture enables the enforcement as well as the anticipation of safety constraints for systems with higher relative degree. The effectiveness of the proposed approach is illustrated on the adaptive cruise control benchmark.

Super-radiance is a striking phenomenon resulting from the collective interaction of emitters with light, and it is fundamentally related to the appearance of long-range correlations between the dipole moments of the emitters. As such, it represents a distinct phase of dissipative many-body systems compared to the case of independent emitters, similarly to how ferromagnetism stands out in magnetic materials in contrast to paramagnetism. In this work we address the conditions under which the transition to super-radiance can occur in ensembles of emitters subject to dephasing and individual decay -- a situation which is particularly relevant to the case of emitters in the solid state. The simplifying assumption of an ensemble of permutationally invariant emitters allows for the efficient solution of both the dissipative dynamics after a pulsed excitation, as well as of the steady state under incoherent pumping. This exact solution allows us to benchmark a truncated cumulant expansion approach, which can give predictions for arbitrarily big system sizes. We show that super-radiance is fundamentally robust to sizable dephasing and non-radiative decay rates, both under a pulsed excitation, as well as under continuous pumping. This robustness is the result of the collective acceleration effect of super-radiant emission with respect to the individual coupling to a local environment. We establish the universal critical scaling laws at the transition between normal radiance and super-radiance; and we show that, in the super-radiant phase, significant finite-size crossovers can be observed before reaching the asymptotic scaling regime. Our results pave the way for future experiments to provide a quantitative characterization of the scaling properties of super-radiance, seen as a distinct non-equilibrium many-body phase in ensembles of incoherently pumped emitters.

Replicator dynamics (RD) is a fundamental model in learning in games, connecting evolutionary game theory and online learning. This paper studies payoff-based higher-order variants of RD represented as a cascade interconnection between an integrator in parallel with an auxiliary linear time-invariant (LTI) system and the softmax mapping. We investigate learnability of Nash equilibria under this Nash-stationary learning rule. First, we revisit recent results that establish convergence to Nash Equilibrium whenever the auxiliary LTI system is strictly passive and prove a converse passivity result: if the auxiliary LTI system is not passive, then there exists a static strictly contractive game whose interior Nash equilibrium is unstable under the closed-loop learning dynamics. Second, we show that there exists a class of games with isolated interior Nash equilibria that cannot be locally asymptotically stabilized by any payoff-based higher-order RD whose auxiliary LTI system is asymptotically stable and strictly proper. Finally, we show that if Nash stationarity (i.e., all Nash equilibria are stationary points of the learning dynamics) is relaxed, then generalized exponential RD (Ex-RD) can locally asymptotically stabilize a logit equilibrium for any continuously differentiable game. The stabilized equilibrium can be viewed as an entropy-regularized approximate Nash equilibrium.

The consensus of continuous-time multi-agent systems with unbounded and heterogeneous constant delays is investigated by combining frequency-domain analysis and algebraic graph theory. Several types of signed Laplacians are constructed to characterize consensusability under delays. The core results are established based on the defined delay-embedded signed Laplacian, where a small-delay link creates a cooperative interaction and a possibly unbounded large-delay link creates an antagonistic interaction between the agents. The dividing line between small and large delays is given by $τ_{ij}=π/2λ_{\max}(\bm{L}_0)$, where $λ_{\max}(\bm{L}_0)$ refers to the maximum eigenvalue of the conventional graph Laplacian. It is proved that the consensusability is preserved if the delay-embedded signed Laplacian is positive semi-definite with a simple zero eigenvalue. Moreover, we derived some consensus conditions in terms of the extended effective resistance which measures the overall coupling between two sets of agents. The obtained results provide new insights into the mechanism of delayed consensus from the interplay between the small-delay-induced cooperativeness and large-delay-induced antagonism in the underlying network topology.

Agentic applications are shifting AI serving from isolated model inference to long-running workloads in which LLMs coordinate tools, environments, and persistent state. However, the system behavior of these workloads---where latency, cost, and bottlenecks arise---remains poorly characterized, leaving serving systems to rely on assumptions built for conventional inference. We present AgentSysBench, a benchmark suite and measurement toolkit with ten representative agentic applications and unified systems-level instrumentation. Across controlled deployments and production traces, we identify six properties that distinguish agentic workloads from conventional LLM serving: (1) execution is heavyweight and stateful, with non-LLM components dominating latency in 5 of 10 applications and sandbox working-set memory peaking at 28 GB per session; (2) applications compose components with heterogeneous resource affinity---GPU-bound inference, memory-bound retrieval, CPU-bound sandboxes---whose task latencies diverge by up to 32x; (3) bottlenecks shift across requests, models, and deployments; (4) production sessions hold state idle for minutes to hours between active steps; (5) a control-plane tax---auxiliary LLM calls and context overhead from tool schemas and observations---crowds out productive compute and context; and (6) production traces from three applications reveal heavy cross-request redundancy in search queries and web fetches, exposing a large caching opportunity. Four design explorations demonstrate that these findings are actionable: task-aware serving reduces latency by 29--40%, communication-aware placement by up to 4.5x, state offloading reduces memory usage by 4.6x, and tool-result caching removes 35.2% of redundant search calls and saves 19.3% of aggregate search latency.

Nanofluidic memristors preserve a record of electrical activity via ion migration and alterations in conductance that depend on the history of the device s state. These characteristics make them suitable for aqueous, energy efficient, and biologically compatible neuromorphic systems. To establish the viability of fluidic memristors for mimicking the brain s dynamic behavior, a more thorough understanding of the memristive materials and the underlying switching processes is required. In this study, we systematically examined a recently introduced memristive device based on inpore chemical reactions, where the combined influence of electrolyte composition and pore architecture on precipitation gated memory remains poorly understood. To address this, we constructed an asymmetric electrochemical system using CaCl2 and phosphate solutions separated by SiNx solid state nanopores. We explored how variations in pH, phosphate concentration, pore geometry, and voltage pulsing regimens affect the electrical characteristics and memristive performance. Comparison of the single pore and the array showed that parallel pores produced smoother pH and concentration dependent hysteresis and pulse responses, whereas the single pore retained larger, nonmonotonic changes.

The rapid advancement of large language models (LLMs) has created a growing asymmetry in cybersecurity, where attack accelerates toward autonomous execution while defense remains predominantly human-intensive. Despite substantial prior work across cyber ranges, AI-driven attack, and AI-driven defense, this asymmetry persists. We trace it to a deeper root cause, that evolution itself has stalled on both sides at three layers. To overcome this, we propose co-evolution as the integrating insight, where attack and defense AI agents autonomously and safely drive each other's evolution through adversarial confrontation. Based on this insight, we present \sysevolve, comprising three co-designed components, \sysfield, \sysspear, and \sysarmor. \sysfield constructs realistic multi-host ranges. \sysspear generates efficient, safe attack schemes. \sysarmor performs real-time, interpretable defense. Together they form a self-driven adversarial loop restoring evolution at all three layers. In evaluation, \sysfield achieves zero-loss collection at 2.1\% overhead and orchestrates 257 CVEs into 1,148 ranges, \sysspear improves attack success by over 25\% over baseline LLMs, and \sysarmor achieves 10--1000$\times$ greater precision than prior systems and detects real APT attacks in production at Huawei and Sangfor. Our evaluation also reveals three findings about LLM agent capabilities. First, multi-step composition and larger topologies expose agent capability gaps hidden by single-step evaluations. Second, the bottleneck lies after initial access in post-compromise state utilization. Third, LLM agents are susceptible to environmental interference. When decoy endpoints are deployed in the range, agent timeouts triple and downstream completion disappears despite the success rates of initial accesses are unchanged.