Physical AI Brief

Current electroencephalography (EEG)-based dream detection relies on power spectral density (PSD) and statistical moment features, achieving a state-of-the-art area under the receiver operating characteristic curve (AUC) of approximately 0.70 on the DREAM database (Wong et al., 2025, Nature Communications). We introduce PHINN-EEG (Persistent Homology Inspired Neural Network for EEG), the first topological time-series framework for dream mentation analysis. Using sliding-window Takens delay embeddings and Vietoris-Rips filtrations on multichannel pre-awakening EEG epochs, we extract Dynamic Betti Curves that characterize the geometric architecture of neural activity, not merely its energy. These topological invariants, combined with topology-conditioned flow matching, are analytically projected to outperform existing PSD and catch22 benchmarks, targeting AUC = 0.82-0.90 on the 1,462-awakening open-access subset of the DREAM database (drawn from a full registry of 3,191 total awakenings from 263 participants across 20 independent laboratories). We further introduce a topology-conditioned rectified flow model for dream-state EEG synthesis-with a spectral-conditioned flow model of comparable feature dimensionality as an additional ablation baseline to isolate the value of topological conditioning specifically-and propose a set of candidate Betti transition archetypes linking topology to phenomenological dream report categories, presented as an exploratory hypothesis space pending empirical validation. If validated, this work represents a paradigm shift from spectral energy to phase-space geometry in neural rare-event detection, with potential future implications for wearable BCI dream monitoring.

In this work, we aim to address the challenge of long-range memory in panoramic world models by exploiting the rotation-equivariant property of omnidirectional representations, where rotation can be treated as an implicit geometric transformation.Building on this insight, we propose PanoWorld, which simplifies camera trajectories into translations via fixed headings for both current-action modeling and long-range memory through Dense Panoramic Ray-Conditioning (DPRC) and Geometry-aware Memory Augmentation (GMA).Then, a three-stage training pipeline is introduced to progressively optimize each component. To better evaluate physical consistency under large-scale spatial variations and diverse illumination conditions, where existing datasets are relatively stable, we construct World360, a large-scale dataset consisting of both real-world video clips collected via panoramic unmanned aerial vehicles and high-quality simulated clips generated by AirSim360.Extensive experiments on World360 demonstrate the effectiveness of PanoWorld, outperforming alternative methods by a large margin.Our models, training code, and dataset will be publicly available. More information can be found on our project page: https://lihaoy-ux.github.io/panoworld-page/.

The rapid progress of large foundation models has been driven predominantly by pretraining on large-scale text corpora. However, many forms of knowledge are conveyed through visual representations, where figures, typeset equations, and page layouts carry rich information that cannot be faithfully or completely captured by text alone. Yet current pretraining approaches discard these visual cues by converting visually rich sources, such as documents and web pages, into plain text for learning language intelligence. This paper challenges the default assumption that language models must be trained on text-only representations and shows that Visual Pretraining is a scalable learner for foundation model intelligence. To this end, we conduct a systematic study of unsupervised visual pretraining paradigms that directly leverage visual documents without text extraction. Across multiple backbones and benchmarks, visual pretraining on the same underlying corpora consistently outperforms text-only pretraining, offering an efficient pathway to scalable language intelligence.

Vision language models (VLMs) have made remarkable progress in visual reasoning during the last decade. Most evaluations have used simple scenes (MS-COCO) that do not showcase complex human interactions or behaviors, only a handful of non-curated human descriptions as a benchmark, and have not focused on understanding the model's error types. Here, we introduce the Complex Social Behavior (CSB) dataset, containing 100 images depicting complex social interactions/behaviors. We analyze the progression of scene descriptions over a decade (2017-2025) of VLMs (four pre-Multimodal Large Language Models, MLLMs, and five MLLMs). We evaluate the accuracy of the models and 20 human descriptions relative to a gold standard on the CSB dataset and on a sample from MS-COCO. We analyzed five visual-cognitive error types: object detection, recognition, hallucination, scene understanding, and spatial dependence. The CSB dataset showed a more pronounced improvement than MS-COCO in scene description accuracy, with pre-MLLMs achieving much lower accuracy than the bottom-ranked human descriptions and MLLMs attaining accuracies similar to the top-ranked human descriptions. We show that MLLMs have eliminated the gap in scene description accuracy between simpler MS-COCO scenes and scenes depicting complex behaviors (CSB). MLLMs have almost eliminated all error types in our tested datasets, except for occasionally relying on different image regions for scene descriptions than humans do (spatial dependence error). We also show that detection, recognition, and hallucination errors have the highest impact on scene description accuracy. Together, our findings provide a more thorough evaluation of how visual language models have advanced over the last decade.

Internet of Things (IoT) systems are inherently vulnerable due to constrained hardware, outdated firmware, and insecure default configurations, creating a need for scalable and adaptive security testing approaches. While recent adoptions of Large Language Model (LLM) agents have demonstrated promise in penetration testing and Capture-the-Flag (CTF) environments, their application to IoT specific vulnerabilities remains unexplored. This paper presents an autonomous multi-agent framework, referred to as Vulnerability EXploitation using AI Agents (VEXAIoT), for vulnerability discovery and exploitation in IoT environments using LLM-based reasoning and offensive security tools. The framework combines a vulnerability detection agent and an attack execution agent to perform reconnaissance, plan attack sequences, and execute exploits against vulnerable IoT services. The system is evaluated in IoTGoat and Metasploitable environments across ten attack scenarios mapped to OWASP IoT vulnerabilities. Experimental results show attack success rate of up to 100% with low token overhead and average execution times under two minutes for most attacks. Across 260 attack executions, VEXAIoT achieves a 95.0% overall success rate, including 94.5% success in IoTGoat and 96.7% success in Metasploitable2. These results demonstrate the potential for LLM-driven agents to automate IoT vulnerability assessment and offensive security workflows in controlled environments

In this work, we present B-spline Policy (BSP), an action representation designed for accelerating robot manipulation policies. Rather than predicting discrete-time action chunks, BSP parameterizes actions as continuous B-spline curves defined by a set of knots and control points. This representation yields smooth, time-continuous trajectories that can be temporally scaled and executed by low-level controllers at higher frequencies and speeds. We show that B-spline-parameterized actions can be seamlessly integrated into standard policy learning pipelines by directly predicting B-spline parameters. Experiments on simulated and real-world tasks demonstrate that BSP significantly reduces task completion time, achieving substantial improvements over baseline methods while maintaining strong success rates. More results: https://b-spline-policy.github.io

Financial anomaly detection suffers from extreme class imbalance, causing traditional single-objective algorithms to exhibit ``fraud collapse'', defaulting to the majority class and failing to balance anomaly interdiction with customer friction. To overcome this without distortive data resampling, we propose the Semantic Pareto-DQN, a multi-objective reinforcement learning framework. Our approach synthesizes heterogeneous transaction features into cohesive natural-language narratives, encoded by large language models, thereby producing a robust, scale-invariant state representation. The agent optimizes a vectorial reward that explicitly decouples financial efficacy, operational friction, and semantic discovery. By mapping the continuous Pareto frontier, the system dynamically navigates the asymmetric costs of missed anomalies versus false positives. Empirical evaluations across E-Commerce fraud and UCI Credit datasets show that semantic Pareto-DQN successfully shatters the zero-recall trap. It achieves superior minority-class recall compared to scalarized baselines, providing an alternative to trade bounded operational friction for financial anomaly discovery.

Quantum information theory (QIT) characterizes the capabilities and fundamental limits of quantum information processing, underpinning quantum communication, computation, and error correction. Formalizing its coding theorems requires connecting finite-block protocols, analytic inequalities, and asymptotic limits within a unified machine-checked framework. Existing developments, however, lack a reusable operational layer that defines codes, error criteria, achievable rates, and capacities independently of their information-theoretic characterizations. In this work, we present LeanQIT, a Lean 4 library for finite-dimensional QIT. It provides composable, kernel-checked interfaces for quantum states and channels, source and channel codes, finite-block performance criteria, hypothesis testing, one-shot quantities, and asymptotic rate constructions. Using this infrastructure, we formalize Schumacher's quantum source-coding theorem, the Holevo--Schumacher--Westmoreland classical-capacity theorem, and the entanglement-assisted classical-capacity theorem together with its strong converse. By separating operational definitions from analytic characterizations and exposing reusable achievability, converse, and asymptotic components, Lean-QIT provides a machine-readable foundation for formal QIT and a compositional knowledge substrate for emerging AI-assisted formalization, automated proof search, and agentic reasoning in quantum information and computation.

Reliable autonomous driving requires full-scene perception that couples foreground objects with dense semantic layout. Recently, 4D millimeter-wave radar has emerged as a robust and affordable sensor, yet its sparse returns make radar-camera fusion necessary for comprehensive scene understanding. Existing radar-camera methods mainly optimize detection, while dual-task systems usually decode boxes and occupancy with limited interaction. To address this gap and advance radar-based multi-task learning, we propose \method, a 4D radar-camera framework for 360$^\circ$ full-scene perception, which models semantic occupancy as a persistent scene state rather than a terminal output. \method{} follows a cross-modal state reasoning paradigm, where the occupancy state is modeled and propagated through stages for coarse-to-fine feature aggregation. Specifically, State-guided BEV Enhancement (SBE) strengthens intra-frame BEV representation, while Doppler-guided Temporal Fusion (DTF) preserves state evidence over longer temporal horizons. Beyond the model, we further extend ManTruckScenes with satellite-map-based generated occupancy labels and pair it with OmniHD-Scenes in a unified cross-dataset detection-and-occupancy protocol. The resulting experiments cover accuracy, robustness, ablation, and efficiency under one radar-camera multi-task evaluation framework. Code and labels will be released upon acceptance.

We present our submission to the QANTA 2026 shared challenge at the ICML 2026 Workshop on Efficient Multimodal Question Answering (EMM-QA). Quanta evaluates multimodal quizbowl systems that answer pyramid-style questions from incrementally revealed text and accompanying images while operating under realistic efficiency constraints. The challenge consists of two distinct tasks: Tossup questions, which require deciding when to answer under uncertainty, and Bonus questions, which emphasize accurate answer selection and human adoption. To address these differing objectives, we develop a task-specific two-agent architecture. Our Tossup agent utilizes a GPT-4o-mini-class model (referred to as GPT-4.1-mini in the competition logs) with confidence-calibrated answering and a domain-specific numeric reasoning policy that reduces overconfident predictions from isolated quantitative clues. Our Bonus agent uses GPT-4o-class model (referred to as GPT-4.1) with leadin-aware reasoning, structured relational reasoning, and multimodal evidence integration to improve exact answer selection. Rather than relying on a retrieval pipeline or model ensembles, our approach emphasizes efficient reasoning policies and confidence calibration within a hosted-only environment. Our system achieved the highest overall leaderboard score of 0.402, including a Tossup score of 0.238 and a Bonus Effect score of 0.164. The results demonstrate that lightweight, task-specific reasoning strategies can provide strong performance on resource-constrained multimodal question answering benchmarks.

As chip complexity increases and time-to-market pressures grow, front-end design has become a critical bottleneck in chip development. Recently, Large Language Models (LLMs) have shown great potential in Electronic Design Automation (EDA). Beyond specification understanding, LLMs show the potential to serve as a unified intelligent interface for hardware description language (HDL) generation, testbench construction, and design space exploration. The rise of agentic AI, represented by pioneering systems such as OpenClaw, offers a strategic roadmap for the next generation EDA. From this perspective, this paper discusses the evolution of EDA from localized assistance to autonomous agentic execution. Then, we review representative advances of LLMs in front-end design, focusing on key tasks such as circuit and testbench generation from a shared specification, as well as design quality improvement in established workflows such as high-level synthesis. Finally, we discuss the key challenges and limitations of integrating LLMs into EDA, and outline future opportunities for advancing LLM-enabled front-end design, offering a systematic perspective for researchers interested in leveraging agentic AI technologies for EDA.

LLM-based multi-agent embodied planning remains impractical due to prohibitively high execution latency. We identify failed actions as the dominant bottleneck, stemming from two core challenges: inaccurate state tracking under partial observability and inefficient coordination that produces redundant or conflicting actions. We introduce Mosaic, a runtime-efficient multi-agent planning framework that addresses both challenges. Mosaic maintains accurate yet lightweight state tracking through agent-centric semantic memory that stores objects in relative coordinates, enabling geometric transformations and coordination. It ensures efficient coordination through Integer Linear Programming that allocates actions at every planning step, enforcing physical feasibility and inter-agent coordination constraints. Across AI2-THOR and search-and-rescue benchmarks, Mosaic achieves 27-32% faster execution, 30-33% fewer LLM calls, 25-31% fewer steps, and 4-10% points higher success rates. These results demonstrate that efficient memory and constraint-guided coordination are critical for scalable, low-latency multi-agent planning.

Enhancing the reasoning capabilities of large language model (LLM) agents requires effective orchestration of diverse expert models and tools. However, existing frameworks typically call APIs based on coarse-grained matching between tasks and the functions of expert models or tools, while overlooking critical factors such as performance variability and cost efficiency among functionally similar alternatives. To address this, we propose Agora, a framework that introduces an incentive-compatible auction mechanism for dynamically allocating tasks to expert models and tools. By treating reasoning steps as tradeable items, Agora enables agents to bid based on their rectified competence-ensuring that critical logic is routed to the most capable solver rather than the most overconfident one. Evaluations across five benchmarks show that Agora improves over matched single-model, routing, and cascade baselines under comparable candidate pools, while exposing a controllable cost-quality trade-off through a single auction parameter.

Precision industrial contact manipulation requires reliable robot policies under pose perturbations and contact-force constraints. Vision-language-action models offer broad generalization but often introduce high inference latency and GPU-memory cost, while vision-action chunking policies are more suitable for real-time industrial control. However, these policies are usually trained by behavior cloning and suffer from distribution shift in contact-rich tasks. This paper proposes PAC-ACT, a reinforcement-learning post-training framework for pretrained Action Chunking Transformer policies. PAC-ACT reformulates policy optimization at the chunk level, constructs an ACT-transferred actor-critic architecture, and introduces a hybrid behavior-prior constraint to preserve the pretrained action distribution during online fine-tuning. Experiments on industrial precision-contact benchmarks show that PAC-ACT improves task success, contact stability, and force safety while retaining low latency and low GPU-memory usage. On the Contour task, PAC-ACT significantly reduces peak contact force and decreases the proportion of force readings above 60 N by 46 times. Sparse-reward ablations further show that the proposed behavior-prior constraint enables effective exploration under randomized initial poses.

Decentralized multi-robot coordination under partial observability remains challenging, especially in communication-free settings where agents must act solely from local sensor observations. Privileged policy distillation provides a promising approach by transferring knowledge from a globally informed oracle to sensor-constrained students. However, in multi-agent systems, the same local observation may correspond to multiple global configurations requiring qualitatively different cooperative actions, making the conditional action distribution inherently multi-modal. Standard deterministic distillation collapses these modes to their mean, often yielding invalid or hesitant actions. To address this issue, we propose CoDiMAD, a three-stage framework that trains a privileged oracle with MAPPO, constructs an offline dataset of local-observation-oracle-action pairs, and distills the oracle into decentralized students parameterized as conditional denoising diffusion probabilistic models. By approximating the conditional oracle-action distribution through the diffusion reverse process, CoDiMAD samples decisive actions from coherent coordination modes rather than averaging across them. Theoretical analysis characterizes the mode-averaging failure of deterministic distillation and the distributional recovery property of diffusion-based distillation. Experiments on three cooperative tasks show that CoDiMAD consistently outperforms direct local MARL and deterministic distillation baselines. The source code will be made publicly available upon acceptance.

The proliferation of agentic AI systems across enterprise and public-sector contexts has outpaced the capacity of general-purpose AI risk frameworks to classify and govern them. In this paper, we introduce the TrustX Agent Risk Classification Framework, a structured, repeatable instrument that can be applied to seven types of agentic AI systems and is grounded in foundational pre-existing AI governance frameworks. At the core of the framework is a twelve-dimension scoring rubric that robustly quantifies the risk. This rubric is combined with other components, such as the GPA + IAT classification model and the five-level autonomy framework derived from existing literature. These inputs produce a three-tier governance output with mapped control recommendations. A specialised Coding Assistant extension is also included to account for nuances specific to this type of agentic AI system. We then use an illustrative example to show our framework in practice. ARC is intended for AI governance practitioners, risk officers, developers, and regulators, and it will regularly undergo iteration as we continue to expand it and make it more robust. The community can access the interactive framework here: https://arc.responsible.ai/

The deployment of large-scale foundation models, such as the Segment Anything Model 3 (SAM 3), promises a transition toward open-vocabulary, training-free computer vision. However, their capacity to generalize out-of-distribution to the complex, top-down geometric structures of Earth Observation imagery remains largely unquantified. Driven by SAM 3's performance disparities in highly specialized domains, we present a comprehensive, multi-task empirical evaluation across remote sensing scene classification, object detection, and instance segmentation under strict zero-shot and one-shot constraints. To achieve this, we introduce a structural adaptation of SAM 3 by repurposing its decoupled binary presence head into a standalone zero-shot classifier. Furthermore, by systematically isolating textual and visual prompt modalities across five configurations, we explicitly diagnose the alignment mechanics within the model's multimodal decoder. Our findings reveal severe cross-modal interference: while visual prompts successfully align the decoder to complex remote sensing geometry, textual prompts inject misaligned, ground-level semantic bias, actively degrading coordinate regression. To benchmark these capabilities without resource-intensive training, we formulate a novel training-free proxy evaluation protocol for Generalized Zero-Shot tasks (scene classification and instance segmentation). Ultimately, our results demonstrate that SAM 3 avoids the overfitting commonly seen in legacy domain-adapted models, achieving high Harmonic Mean scores in segmentation tasks. However, it remains fundamentally constrained by sub-pixel resolution limits and overhead semantic blind spots, charting a definitive mandate for parameter-efficient geospatial fine-tuning of its multimodal decoder.

Pre-demolition assessment, the regulated audit process at the heart of urban mining, is an information process in which AI support must serve qualified auditors who remain accountable for the decisions taken. The relevant unit of value is not prediction accuracy alone, but the defensibility of the supported decisions: their legibility, plausibility, sourcing, and contestability. Explainable AI techniques and domain knowledge graphs each address parts of this requirement, and existing taxonomies have catalogued their integration. The literature is descriptively rich but structurally under-specified: what remains less developed is a structural account of why specific integrations produce artefacts neither resource can provide alone. This paper offers a complementarity-theoretic interpretation grounded in the IS resource-based tradition. We propose four consolidated KG-XAI integration modes (Lifting, Constraining, Typing, and Revising), each defined as a typed operation over XAI artefacts and knowledge-graph substrate structures. Each mode unlocks a distinct property of defensibility and contributes to the kind of regulatory artefact pre-demolition assessment demands. A fire-door example from the urban-mining process illustrates the modes using the W3C Linked Building Data stack and valuation extensions.

Decision-making is posing an increasingly formidable challenge to investors because of the growing number of alternatives available in financial markets. A hot area of research over the past few decades has been portfolio optimization that seeks to determine how much an investor should invest in which asset. Introducing real-world conditions to the optimization model turns the problem into an NP-hard one for whose solution exact methods become inefficient; hence, researchers have turned to evolutionary algorithms to approximate solutions. In this paper, strengthening strategies are presented for multi-objective evolutionary algorithms that can provide a faster convergence rate and extensive search ability in the portfolio optimization problem under the cardinality constraint. To implement those features, a unique solution representation, a novel operator, and new repair mechanisms are introduced for solving the aforementioned problem in which lower and upper limits are set on the number of assets in the portfolio. For this purpose, new mating strategies along with the aforesaid package are implemented in well-known multi-objective evolutionary algorithms to solve the problem. The customized algorithms are subsequently tested against traditional ones using well-known market indices as benchmarks. Results indicate that the proposed strategy not only provides better approximations but also converges faster as well at no loss of performance with an increasing number of assets in the market.

Modern AI systems are increasingly being evaluated for their ability to reason, code, prove theorems, use tools, and long-horizon research tasks. These are powerful capabilities, but they share a structural limitation: the representational frame within which the model operates, including its conceptual vocabulary, the space of admissible solutions it can search, and the criteria by which success is evaluated, is typically fixed and supplied in advance. This paper argues that building stronger intelligent systems capable of open-ended innovation requires additional classes of operations: the creation, stabilization, and reuse of new representational primitives, which alter the space being searched rather than simply searching within it. We characterize the distance between current AI systems and genuinely open-ended intelligence through two gaps. The first is the vocabulary gap, the difficulty of inventing and stabilizing new representational primitives rather than merely recombining existing ones. The second is the verifier gap, the difficulty of judging the value of a new primitive when its full payoff may be visible only after future reuse. We interpret both gaps through a unified framework of intelligence as cognitive discrepancy reduction. By viewing intelligent behaviors as a sequence of cognitive transformations, we distinguish intra-space transformations which operate within a fixed representational frame, from generative transformations which may modify the frame itself. On this basis, we propose a ladder of innovation autonomy and outline several directions for advancing open-ended AI, including objectives that reward useful representational change, persistent memory architectures for invented primitives, and adaptive verification mechanisms capable of evolving alongside the representations they evaluate.

Fine grain control and positioning of autonomous underwater vehicles (AUVs) is critical for sampling, maintenance, and survey applications. Traditional control methods for AUVs are labor intensive and are not robust to changes in the vehicle configuration or environmental conditions. Reinforcement learning (RL) promises rapid controller development while handling a range of deployment parameters via domain randomization (DR). However, DR is still limited by the capacity of the underlying simulation to model real physics. In particular, drag physics are difficult to model and are a large contributor to sim-to-real gaps. Meanwhile, computational fluid dynamics (CFD) provides high fidelity drag models but is challenging to leverage within reinforcement learning frameworks due to its computational overhead. Thus, in this paper we exploit the idea of training surrogate approximations of CFD models of a given vehicle, enabling fast inference within RL pipelines. We are the first to successfully deploy a zero-shot RL policy on a 6-DOF AUV in which policy training is performed on surrogate drag models (SDMs) trained on CFD data. We find 31% lower energy usage compared to a controller using simplified physics while traversing between waypoints 11% faster with 19% less error. Our SDM based RL controller better predicts zero-shot transfer and is more robust across reward shaping design choices. When using DR to complete a task with perturbed parameters, we find that the CFD policy is the only controller that successfully transfers. The policies are evaluated in a controlled tank environment and in the field providing extensive testing of the policies' capabilities.

Aerial manipulation with multirotor platforms enables physical interaction in complex environments, but rotor-induced airflow remains a critical limitation for tasks involving airflow-sensitive targets or surroundings. This paper presents an optimization-based design framework for modular aerial manipulators that jointly considers task wrench feasibility, end-effector placement, and airflow exposure constraints. We first introduce a novel categorization of target-side airflow tolerance and formulate the corresponding exposure requirements as geometric constraints. To efficiently model rotor-induced airflow, we introduce a compact cone-sphere envelope that approximates the spreading structure of a quadrotor's airflow while preserving computational tractability for optimization. Building on this formulation, we propose a reconfiguration optimization that adapts a modular aerial manipulator to diverse task wrench requirements while enforcing both target-side airflow exposure and intra-platform airflow interference constraints. Unlike prior designs that assume a fixed end-effector location, the proposed framework optimizes the end-effector placement together with the platform configuration. Scalability experiments and ablation studies validate the effectiveness of the proposed framework.

Despite strong performance on many multimodal tasks, vision-language models (VLMs) still struggle with basic object counting. We investigate whether this reflects missing internal knowledge or a gap between internal representations and verbalized outputs. Training simple probes on activations from four VLMs across five counting datasets reveals that nonlinear probes can reliably detect counting errors, suggesting that VLMs often encode the correct count even when they output the wrong answer. SVCCA analysis shows that probes trained on ground-truth counts and probes trained on model outputs occupy a partially shared activation subspace but read out along misaligned directions. We further validate our findings using a causal steering intervention, proving that strengthening the direction of count-identified probes does improve model counting performance. Motivated by this result, we propose a detector-guided self-correction method that selectively re-prompts the model only when an internal error detector predicts failure. This simple inference-time intervention improves counting accuracy by up to 15.6 absolute percentage points, without any parameter updates. Our results establish activation-based error probing as both a practical tool for improving VLM counting and a mechanistic lens on the gap between internal knowledge and model outputs.

Self-supervised pretrained foundation models (FM) have shown early promise for non-invasive electroencephalogram (EEG) decoding applications. Many recent large-scale models converged on the approach of tokenizing raw EEG followed by masked reconstruction pretraining. However, this recipe has been shown to be suboptimal for data, like EEG, with high noise amplitude and information confined to limited dimensions such as narrow frequency bands. Building on this insight, we develop a novel contrastive-pretrained EEG model with multiscale temporal convolution input layers and Transformer encoder blocks (CoCoT). CoCoT matches or beats state-of-the-art reconstruction-pretrained EEG models on extensive benchmark decoding tasks with heterogeneous electrode configurations. Furthermore, CoCoT trained from scratch outperforms previous single-task decoding models and even rivals pretrained models, showcasing the architecture's flexibility and data efficiency. Through systematic ablations, including model architecture and pretraining objective, we demonstrate the viability of contrastive learning for building EEG FMs while suggesting key architectural design considerations, prompting further investigations in alternative large-scale pretraining strategies.

Time series forecasting requires models to capture diverse, often mutually exclusive, temporal dynamics, from smooth trend continuation to nonstationary drift and strict phase-aligned recurrence. While recent deep learning models have improved accuracy, they typically force these diverse patterns through a single computational backbone governed by fixed algorithmic inductive biases (e.g., self-attention or spectral filtering). This single-mechanism approach often struggles with the profound heterogeneity of real-world series, where different variables and forecast horizons necessitate fundamentally different predictive treatments. To address this, we propose GatedLinear: a lightweight framework that frames forecasting as the adaptive routing of complementary linear bases. GatedLinear leverages a pool of three specialized mechanisms: a global trend-seasonal basis for smooth projection, a difference-based incremental basis for nonstationary drift, and a phase-aligned recurrence basis for explicit cyclic reuse. To dynamically orchestrate these distinct behaviors, we introduce a Tri-Factorized Fusion Gate that disentangles routing decisions into channel-specific preferences, horizon-aware offsets, and phase-indexed biases derived from known future time marks. This design allows the model to perform highly granular, point-wise soft routing across different predictive regimes without stacking computationally heavy neural modules. Experiments on standard benchmarks show that our method achieves state-of-the-art or highly competitive accuracy against recent complex foundational models, while offering explicitly interpretable routing patterns and operating with a substantially smaller parameter footprint.

We introduce Freya-TTS, a compact, tokenizer-free, Turkish-first text-to-speech model designed for highly reliable and efficient conversational synthesis. Freya-TTS is a 183.2M-parameter non-autoregressive conditional flow-matching Diffusion Transformer (DiT) that operates in the frozen continuous latent space of AudioVAE2 (16 kHz encode, 48 kHz decode), allowing the model to focus its capacity on text-to-latent mapping while inheriting high-quality 48 kHz reconstruction. We advance the framework along three key dimensions: (1) rule-free end-to-end modeling from a 92-symbol Turkish character vocabulary without a phonemizer, grapheme-to-phoneme frontend, or discrete speech tokenizer; (2) non-autoregressive parallel denoising, which predicts the entire latent sequence simultaneously over a predicted duration; and (3) a production-oriented two-stage post-training recipe consisting of single-speaker voice locking and short-utterance coverage, improving speaker consistency and robustness on short inputs. On the Freya-TR-Eval benchmark, Freya-TTS achieves a band-matched word error rate (WER) of 8.0% and character error rate (CER) of 3.0%, outperforming substantially larger open-source systems while using a fraction of their parameters. The model achieves a real-time factor of 0.11 on consumer GPUs and runs faster than real time on a laptop CPU, making it well suited for resource-constrained edge deployment. We release the model weights, training and inference code, and evaluation benchmark under the Apache-2.0 license.

The emergence of metaverse platforms has created virtual economies that introduce new challenges related to fraud, bot activity, and illicit financial behavior. Despite growing interest in trustworthy metaverse analytics, existing datasets typically focus on user behavior, authentication, or financial transactions in isolation, limiting the development and reproducible evaluation of multimodal fraud detection methods. To address this gap, we present TSAI-MetaFraud, a multimodal, multi-task benchmark dataset for fraud analytics in virtual economies. TSAI-MetaFraud integrates behavioral, transactional, and graph-structured information while incorporating realistic fraud and automated bot scenarios. We define benchmark tasks including transaction fraud detection, cross-modal node classification, temporal link prediction, and weakly supervised fraud detection, and provide baseline evaluations using machine learning models and graph neural networks. By jointly capturing behavioral activity, financial interactions, and relational structure within a unified virtual economy, TSAI-MetaFraud provides a benchmark for advancing multimodal learning, graph mining, fraud analytics, and trustworthy AI in emerging metaverse ecosystems.

Does every cancer patient truly need a complete diagnostic workup for accurate survival prediction? In multimodal clinical oncology, diagnostic modalities follow a clinically mandated order of escalating burden -- from demographics collected at intake to genomic profiling requiring specialized tissue analysis. Current multimodal survival methods either assume all modalities are available or passively handle missing data, but none actively reason about whether acquiring the next modality is justified for a given patient along this ordered workflow. We formulate this as a sequential decision problem and propose SAGEAgent (Sequential Acquisition Guided by Experience), a self-evolving LLM-based clinical agent that decides which diagnostic modalities to acquire for each patient, balancing predictive accuracy against clinical invasiveness. SAGEAgent reasons about each patient's evolving diagnostic state through clinical tools that translate numerical predictions into text, an episodic memory that retrieves similar past cases, and a semantic memory that accumulates reusable decision patterns from experience. Experiments on a glioma cohort combining TCGA-LGG, TCGA-GBM, and BraTS with four diagnostic modalities demonstrate that SAGEAgent achieves competitive survival prediction accuracy while reducing average acquisition burden by 55%.

Vision-Language Models (VLMs) are the perceptual backbone of embodied AI, but their energy footprint on edge hardware remains poorly understood. Existing efficiency efforts focus predominantly on reducing visual tokens, implicitly treating visual processing as the dominant energy cost. We overturn this implicit assumption through the first systematic energy profiling of on-device VLM inference, spanning five models across three architecture families, four input resolutions, and two hardware platforms (NVIDIA RTX 3070 and Jetson Orin NX). Our analysis yields three findings. First, average inference power is a model-intrinsic constant, invariant to input resolution, image complexity, and prompt type, with less than 5% variation across all conditions. This means that all energy variation across inputs must arise from variation in inference time, not from variation in power draw. Second, each output token costs 11 to 39x more wall-clock time than each input token due to the compute-bound and memory-bound asymmetry between prefill and decode, making output token count the dominant driver of both latency and energy. Third, image complexity, measured by the number of objects in an image, induces up to 4.1x energy differences at identical resolution. This variation arises not from increased visual processing cost, but from differences in output length. These findings expose a fundamental limitation of visual token pruning: even removing all visual tokens saves at most 10% of total energy for fixed-token models. Across models spanning 1 billion to 8 billion parameters, controlling output length saves up to 97% of total energy, with the energy dominance of decoding growing stronger at larger model scale. In short, the true energy bottleneck in edge VLM inference is not what the model sees, but how much it says.

We present DemoBridge, an toolkit that turns a single-view RGB stereo recording of a human hand demonstration into an executable, physics-validated robot-arm trajectory. Retargeting across the embodiment gap is hard. A robot arm reaches a target with a long, articulated body whose links carry far more collision volume than a hand. Solving inverse kinematics for the mapped end-effector pose often yields no collision-free solution, and a trajectory imposes this at every waypoint. A single view adds noise, leaving the demonstrated reference inaccurate. At the core of DemoBridge is a single collision-aware planner. It optimizes the whole joint trajectory at once, reasoning jointly over alternative grasp poses, whole-arm and grasped-object collision, and fidelity to the demonstrated path. A physics simulator runs in the loop. It validates each phase as it is produced and backtracks on failure, so a demonstration that cannot be reproduced as given is re-planned rather than discarded. The resulting action sequence is dynamically stable and faithful to the demonstrated manipulation. It also doubles as a ready-to-use simulation rollout for policy learning. Grasp timing is inferred automatically, and the perception backends, robot, and pipeline stages are swappable from configuration. We evaluate whole-pipeline retargeting on three real-demonstration tasks and the planner on a controlled synthetic benchmark. Our code is available at https://gitlab.kuleuven.be/u0123974/demo-bridge/ .

Robotic manipulation tasks often require simultaneous reasoning over motion and contact forces, yet most Learning from Demonstration (LfD) methods model only spatial trajectories and neglect force interactions with the environment. This limitation reduces robustness and can lead to unsafe or inconsistent task reproduction in force-constrained settings. We propose a novel one-shot multimodal LfD framework for the segmentation, encoding, and reproduction of force-inclusive demonstrations. First, we introduce a multimodal probabilistic segmentation method that adaptively weighs spatial and force modalities over time, enabling the automatic extraction of force-aware motion primitives. Second, we extend the elastic maps representation to incorporate external force constraints during skill encoding and formulate a convex optimization procedure for learning force-consistent trajectory models. The resulting skills reproduce both motion and contact characteristics from a single demonstration while promoting safer execution by accounting for demonstrated force profiles. We validate our approach on five real-world manipulation tasks across two distinct force-sensing configurations: wrist force sensing on a UR5e with a Robotiq 2f-85 gripper and finger force sensing on a Kinova Gen3 with an Openhand Model O gripper. Experimental results demonstrate robust multimodal segmentation, accurate force-aware reproduction, and cross-platform generality.

Large language model (LLM) coding agents are increasingly deployed to autonomously perform software engineering tasks in terminal-based environments, making their reliability a growing concern. Existing empirical studies investigate why coding agents fail, yet they largely treat failure as a final outcome rather than a temporal process, providing limited insight into how failures emerge, evolve, and become unrecoverable. We present the first large-scale empirical study of CLI coding-agent failure trajectories, introducing a process-oriented framework that analyzes failure through its onset, evolution, and recovery across execution trajectories. We first collect 3,843 execution trajectories generated by seven frontier models across three coding-agent scaffolds (OpenHands, MiniSWE, and Terminus2) on Terminal-Bench, then carefully filter them to obtain 1,794 complete and valid trajectories for manual annotation (over 63,000 execution steps), from which we derive 14 findings spanning failure occurrence, root causes, recovery, and cross-system consistency. Our findings show that coding-agent failures are predominantly driven by epistemic errors, typically begin within the first few execution steps, and often remain hidden until recovery is no longer possible, suggesting that improving coding-agent reliability requires earlier validation and intervention rather than relying solely on final-outcome evaluation.

A fundamental challenge in 3D reconstruction and robotic localization is co-visibility: determining which image pairs share overlapping visible surfaces, particularly in scenarios with minimal overlap. We demonstrate that VGGT implicitly encodes co-visibility as an emergent behavior: without any supervision for this task, its internal representations exhibit a clear hierarchical structure mirroring that of large language models, i.e. early layers build a 3D-aware scene representation, while late layers act as dedicated co-visibility reasoners. In particular, we identify layer L17 as a negative anchor that consistently routes non-co-visible pairs for this backbone, regardless of the evaluation setting, providing task-grounded evidence of layer specialization in a geometry-grounded foundation model. Building on this, we introduce Co-VGGT, which freezes VGGT and trains only a lightweight layer-wise mixture-of-experts head (less than 7.5M parameters) to classify co-visibility from RGB alone, treating each layer as a specialized expert whose geometric abstraction is adaptively weighted per input pair. On the Co-VisiON benchmark, Co-VGGT surpasses the human annotation baseline and improves over prior work by more than 25% pairwise and 10% multiview. Pairwise predictions are well-calibrated (ECE=0.030), enabling direct use as edge weights in visibility graphs for downstream SfM and SLAM pipelines without post-hoc correction. Code and data are available.

Explaining machine-learning models is increasingly important for decision-making and consumer trust, yet it is widely believed to come at a cost: existing Explainable AI (XAI) methods suffer from a persistent accuracy-explainability trade-off. We argue that this trade-off is not fundamental, but an artifact of treating explanation and prediction as separate objectives; when properly coupled, they become complementary, so that equipping a model to explain itself improves, rather than degrades, its accuracy. We introduce the Rashomon Explanation paradigm, which builds a set of faithful, prediction-guiding explanations rather than a single one, and prove that this set is generally non-empty and that explanation fidelity bounds the performance of the models it guides. To explore this set, we propose RashomonLLM, an Explanation-Prediction-Reflection agentic workflow that generates explanations in natural language by iteratively aligning them with predictions, and we prove it converges and recovers the full set. Across customer-churn classification, clinical survival regression, and industrial click-through prediction on large-scale live-streaming logs, RashomonLLM significantly outperforms state-of-the-art prediction and XAI baselines on both accuracy and explanation quality, with gains driven by explanation fidelity and robust to distribution shifts, temporal splits, and seeds. Our framework thus advances business performance while laying the groundwork for consumer trust.

Agentic LLM systems that generate code through multi-turn tool use face a fundamental context problem: each session starts from zero, discarding the configuration choices, domain constraints, data schemas, and tool-use patterns that made previous sessions productive. Naively persisting entire conversation histories is token-inefficient and counterproductive: irrelevant context degrades generation quality. We introduce shared selective persistent memory, an architecture that identifies and retains four categories of reusable context (task specifications, data schemas, tool configurations, and output constraints) while discarding session-specific reasoning traces. Crucially, this memory is shared: workspaces encapsulating selective memory can be transferred across users with role-based access control, enabling collaborative reuse without redundant specification. We implement it in a deployed collaborative workspace platform where LLM agents produce, edit, and maintain git-versioned artifacts (dashboards, reports, and data-driven documents) from heterogeneous sources (CSV, SQL, REST APIs, and MCP servers). A complementary zero-token data refresh mechanism decouples generated programs from runtime data, enabling artifact reuse without re-invocation. Across three enterprise scenarios, shared selective persistent memory achieves 96% task completion (vs. 79% without memory and 71% with full history). Zero-token refresh eliminates LLM re-invocation for recurring updates (14x task-time reduction), while summary-driven generation cuts per-invocation token cost by 97x versus raw data injection. A replication on four public datasets confirms generalizability, with zero-token refresh succeeding in 12/12 trials. Notably, naive full-history persistence actively degrades completion by biasing the agent with stale traces, while selective memory outperforms both extremes.

High-quality visual representation is a long-standing pursuit in computer vision. In the context of multimodal LLMs (MLLMs), feeding higher-resolution images can produce more fine-grained visual tokens. However, it introduces additional computational and design complexity, due to multiple forward passes and post-processing of increased tokens. Before simply adopting a higher resolution, have we truly unlocked the model's full perception capability at a standard resolution? Therefore, we study an interesting problem: how to achieve fine visual perception under lower cost without larger images. We present SigLIP-HD in this work. The core is a highly simple fine-to-coarse supervision design. We enforce the coarse feature of a mid-resolution image to mimic the fine-grained feature of its high-resolution version. We build this framework on the advanced SigLIP 2 model. Our final model produces better visual tokens at exactly the same inference budget. It is validated on extensive MLLM benchmarks and consistently delivers stronger results than our baseline model, especially on OCR-related tasks.

Large language models (LLMs) have shown increasing promise in solving open problems in mathematics. However, their performance can be further improved through agentic workflows tailored to real-world mathematical practice. To this end, we introduce ProofCouncil, a mathematical agent that is designed to tackle open problems using an author-critic architecture. ProofCouncil served as a submission to the second batch of FirstProof, a challenge consisting of 10 real-world mathematical problems that agents must solve autonomously. Its submissions for 6 of the 10 problems were judged by the referees to be correct up to at most minor revisions, showing the best performance among participating teams. We also evaluate ProofCouncil on 30 open problems collected from mathematical researchers. Among the 21 solutions that received human feedback, 5 were judged completely correct, 2 more were judged promising pending final verification, and a further 8 contained useful partial progress. In this short paper, we describe the development of ProofCouncil and the agent-building library used to create it, which we release as open source to the community.

Universal machine learning interatomic potentials (uMLIPs) bridge quantum-mechanical accuracy and large-scale molecular dynamics, but the cost of high-accuracy calculations such as r$^2$SCAN limits training to datasets that remain small relative to the open materials space. Strong average benchmark performance also does not guarantee reliable energy--force predictions for every structure. We propose Adaptive Multi-Teacher Routing (ATR), which reformulates high-fidelity data construction as a structure-wise decision problem under uncertainty. Using a small set of real r$^2$SCAN labels, ATR calibrates multiple pretrained uMLIP teachers and combines structural descriptors, teacher identity, and inter-teacher disagreement to estimate the reliability of each structure--teacher pair. It selects high-confidence predictions for pseudo-label generation and rejects structures for which no teacher is sufficiently reliable. With real r$^2$SCAN labels for only 0.2\% of candidate structures, ATR distils 2.89 million traceable r$^2$SCAN-level pseudo-labels for pretraining. On held-out r$^2$SCAN structures and the MP-r$^2$SCAN benchmark, a lightweight CHGNet trained on the ATR-generated dataset consistently outperforms the baseline and non-routed controls. Finite-temperature molecular dynamics further shows that ATR improves dynamical robustness across multiple material systems, maintaining stable trajectories where baseline simulations undergo catastrophic structural collapse. These results establish active rejection as an effective mechanism for converting multiple pretrained uMLIPs into a scalable and reliable data-construction system for high-fidelity uMLIPs.

3D scene graphs provide a hierarchical abstraction of environments by encoding spatial entities, such as objects and places, and their relationships. However, existing scene graph systems model object geometry coarsely, relying on partial point clouds or class-level CAD templates, which limits instance-specific shape detail. This paper presents Hydra++, a system-level investigation into how learning-based object shape estimators can be integrated into a hierarchical 3D scene graph pipeline. Hydra++ incorporates category-agnostic shape estimation and a reprojection-mask consistency check to reject degenerate predictions from partial observations or imprecise segmentation. In its default CRISP-based configuration, Hydra++ performs online scene graph construction; slower estimators such as SAM3D are evaluated as modular alternatives to demonstrate generalization-latency trade-offs. Furthermore, to address the challenges of sparse and noisy depth measurements in outdoor environments, Hydra++ supports a hybrid LiDAR-camera configuration for large-scale operation, improving scene-level reconstruction quality. Experiments in both simulation and real-world outdoor campus scenarios demonstrate that Hydra++ improves object- and scene-level reconstruction quality. Project page is available at https://hydra-plusplus.github.io/.

Long-term animal re-identification (ReID) must remain robust to gradual morphological evolution and seasonal appearance shifts. Although recent vision-language models provide strong pretrained visual representations, adapting them to longitudinal ecological settings remains challenging, particularly under identity and temporal distribution shifts. We present a parameter-efficient CLIP adaptation framework for animal ReID and introduce a continuous metadata-conditioning mechanism that incorporates numerical attributes directly into the prompt representation during training. While low-rank visual adaptation, prompt-based supervision, and cross-modal alignment provide the adaptation framework, the proposed metadata-conditioning strategy constitutes the primary methodological contribution. By preserving the continuous structure of numerical metadata rather than discretizing it into textual categories, the proposed approach enables smooth modulation of the embedding space during training while maintaining a purely visual inference pipeline. Experiments on a seven-year longitudinal fish dataset and multiple wildlife benchmarks demonstrate improved performance under closed-set, open-set, and time-aware evaluation protocols. The results demonstrate that continuous metadata conditioning improves robustness to longitudinal appearance variation and temporal distribution shifts, while parameter-efficient adaptation enables a purely visual inference pipeline without requiring metadata at test time. Code and evaluation splits can be found at: https://github.com/AnilOsmanTur/MetaPrompt-ReID.

Test-time scaling (TTS) reliably improves reasoning in large language models, but whether it transfers to small open vision-language models remains unclear. We examine this on EXAMS-V, a multilingual visual multiple-choice benchmark, comparing self-consistency, describe-then-reason with PRM-guided beam search, and two post-hoc selectors across Qwen2.5-VL-7B-Instruct and Qwen3.5-4B. What matters is the conditions under which TTS runs, not the search or verification machinery. The largest factor is parseability: an early prompt format left many chains reasoning correctly yet never committing to an answer letter, which a standard answer cue and a guided repair step largely remove. A larger decoding budget removes the rest: raising the per-chain token limit from 1k to 2k recovers 3.7 pp, whereas sampling more chains (8 to 16) adds only 0.15 pp. Once chains have room to finish, elaborate methods contribute little: PRM-guided beam search trails plain self-consistency by 0.39 pp at over eight times the cost, and neither a training-free generative critic nor a trained multimodal PRM beats majority vote across both policies. The largest gain comes instead from the policy model itself (+11.4 pp). Our best configuration reaches 84.1% on the held-out ImageCLEF 2026 test split, ranking first on the Visual MCQ leaderboard.

We present Soofi S 30B-A3B, a sovereign, open-source Mixture-of-Experts (MoE) hybrid Mamba Transformer foundation model for German and English. Its hybrid design activates only 3B of 30B parameters per token and keeps the inference cache near-constant as context grows, giving it a decisive throughput advantage over dense models for long-context, high-concurrency deployment. Pretrained on roughly 27 trillion tokens with deliberately up-weighted German, Soofi S matches dense 14 to 27B models on aggregate English and German benchmarks while achieving the best code aggregates in both languages among 17 open base models, and outperforms every European sovereign baseline in our comparison, including ones far larger in active parameters. Among fully open models, Soofi S obtains the highest English and German evaluation scores, ahead of Olmo 3 32B and Apertus 70B. Soofi S was built end-to-end on the German Industrial AI Cloud, a sovereign HPC scale AI infrastructure operated by Deutsche Telekom in Munich. Soofi S will be released under highly permissive, open-access terms: weights, selected intermediate checkpoints, full per-source data accounting, hyperparameters, and training and evaluation code. Where source licenses permit, data-construction artifacts are released under permissive licenses; commercially licensed sources are documented with aggregate statistics and exact mixture accounting.

Cooperative multi-agent reinforcement learning is well suited to problems with large parameter spaces and exploitable local structure, such as the tuning of electrostatically-defined quantum-dot arrays. However, if parameter cross-talk is strong, a non-stationary environment from the perspective of any individual agent can destabilize learning - the same effect that plagues manual tuning of such systems. We propose using a factored representation of the action space, learned online, to decouple agents and minimize their interference. Our framework, QADAPT, uses this factorization to efficiently learn shared policies based on local measurements and rewards. With this modular strategy, we achieve zero-shot generalization to unseen quantum device sizes and maintain an approximately constant number of convergence steps to reach target regimes. This work provides a scalable route toward the rapid calibration of large-scale quantum processors.

Long-context processing has become increasingly important for large language models (LLMs), but simply extending the context window does not guarantee effective utilization of long inputs. As input length grows, accuracy often degrades, indicating that models still struggle to identify and use the evidence most relevant to a question. A promising way to improve long-context utilization is test-time training (TTT), which treats the test context as a training example for instance-specific parameter adaptation. However, applying TTT to the entire long context is prohibitively expensive, while adapting on randomly sampled spans introduces severe noise. Because most spans in a long context are irrelevant to the specific question, training on them may even degrade the base model's performance. Our preliminary study shows that TTT is highly sensitive to training-span quality: on LongBench-v2, TTT on randomly sampled spans hurts performance, whereas TTT on oracle spans substantially improves it. Motivated by this, we propose a simple method, Self-Guided TTT (S-TTT): before adaptation, the model identifies the evidence spans it should learn from, and the standard language-modeling training objective is applied only to those selected spans. On two challenging long-context reasoning benchmarks, LongBench-v2 and LongBench-Pro, S-TTT improves accuracy for both Qwen3-4B-Thinking-2507 and Llama-3.1-8B-Instruct, achieving up to a 15% relative improvement.

Motivation: Rare disease (RD) diagnosis is frequently delayed due to the similarities in symptoms to common disease variants. Machine Learning Algorithms applied to Electronic Health Records show promise for accelerating the diagnosis; however, legal and privacy concerns pose significant barriers. To address these issues, Synthetic Data Generation is an alternative method for obtaining Electronic Health Records and can be applied with any Machine Learning algorithm for benchmarking and development purposes. Despite the availability of Synthetic Data Generation algorithms, support for generating a subset of patients that differ in a definable degree from the majority to simulate patients with RD is often lacking. Results: We present SYNRARE, a graphical user interface based on the Synthea framework that enables easier modification and generation of synthetic Electronic Health Records of RD patients, which differ only to a definable degree from patients with common diseases, thereby enabling the benchmarking and testing of algorithms under controlled technical conditions. SYNRARE enables researchers to rapidly benchmark their Machine Learning algorithms across any scenario. Availability and implementation: SYNRARE, including detailed instructions for installing, is available at https://gitlab.sdu.dk/screen4care/synrare.

Worldbuilding, the construction of coherent fictional worlds, is a foundational task in game design and literary creation. Large Language Models (LLMs) offer new possibilities for automated content generation, but their application to worldbuilding faces three challenges: context explosion that grows linearly with the building process, the tension between creative diversity and content consistency, and the absence of automated quality assurance. This paper presents AutoWorldBuilder, a multi-agent collaborative system that addresses these challenges through five integrated components: a structured concept network with conflict detection; a DAG-based hybrid batch scheduler that groups tasks by semantic locality; a four-layer context compression mechanism achieving approximately 90% token reduction; an iterative review system with specialized Auditor agents that improves proposal pass rates from 42% to over 85%; and a skill-driven agent architecture supporting zero-code extension with differentiated temperature configuration. Two experiments across 20 diverse worldbuilding tasks, using GPT-OSS 120B and DeepSeek v3.2 as LLM backends, demonstrate a 95.0% success rate. The system generated 56-103 self-consistent concepts per world in 18-31 minutes with zero-conflict delivery. The architectural patterns validated here, including layer-as-budget compression, semantic-locality scheduling, and separation of generation and review, transfer to the broader class of knowledge-intensive, multi-agent LLM applications.

We introduce a novel method for both partial and full optimization of the connections in deep differentiable logic gate networks (LGNs) and lookup table networks (LUTNs). Our training method utilizes a probability distribution over a set of connections per gate/lookup table (LUT) input pin, selecting the connection with highest merit, all whilst the optimal gate types or LUT-entries are learned in parallel. We show that the connection-optimized LGNs outperform standard fixed-connection LGNs on the Yin-Yang, MNIST Handwritten Digits and Fashion-MNIST benchmarks, while requiring only a fraction of the number of logic gates. We achieve 98.92% on the MNIST dataset with two layers of 8000 gates. With only one layer of 8000 gates, we obtain 98.45%, showing that our method requires almost 50 times fewer gates compared to fixed-connection LGNs. Training stability up to ten layers has been ensured by employing a high learning rate, straight-through estimators and trimming constant-output gate types. Additionally, we present a LUT neuron description that enables stable training with backpropagation, tested up to 6-layer deep networks. The model requires four times fewer trainable parameters and still achieves a higher accuracy compared to the fixed-connection LGN training algorithm. Our connection-training algorithm also works well for the LUTNs, achieving an accuracy of 98.88% for two layers of 2000 6-input LUTs.

The growing adoption of large language model-based agents within operating system workflows has increased the importance of energy-efficient inference on laptop-class systems-on-chip (SoCs). While cloud offloading remains common, it introduces reliability and privacy concerns that are particularly problematic for agentic workloads. Recent laptop SoCs, therefore, incorporate neural processing engines (NPUs) optimized for energy efficiency; however, effectively mapping attention mechanisms onto NPUs remains challenging due to architectural diversity and explicit data-movement programming models. In this work, we present STEEL, the first open-source implementation of FlashAttention targeting XDNA-like NPUs. STEEL introduces a dataflow formulation of prefill attention, enabling efficient exploitation of spatial parallelism and on-chip memory. Furthermore, STEEL addresses the load imbalance induced by the causal mask by leveraging a sparsity-aware pipeline placement onto the NPU array, reducing synchronization overhead and improving utilization. We evaluate STEEL on the AMD Ryzen AI 9 HX 370 SoC and compare its performance against optimized CPU and GPU implementations. Experimental results show that STEEL reduces energy consumption by an average of 9.17x and 1.75x relative to CPU and GPU baselines, respectively. On XDNA 1, STEEL achieves an average 9.6x latency reduction over the prior state of the art, and delivers a 22.8x speedup on average compared to a layer-by-layer attention implementation on XDNA 2.

We study an adversarial bandit problem for entanglement-based quantum-network routing over a modest graph corpus. Alice selects an end-to-end repeater route for an Ekert-91 protocol (E91) representing her move, while Eve selects an attack surface, either edge intercept--resend or repeater memory degradation. Payoffs are drawn from cached SeQUeNCe-simulated E91 transcripts, and Alice accepts a turn when the finite-sample statistic violates the Clauser-Horne-Shimony-Holt (CHSH) bound. Performing adversarial co-learning across 50 structured topologies, we find that learned retention tracks a full-matrix minimax reference closely (Pearson $r=0.99$): under a one-surface Eve action model, bottleneck families have zero retention, while non-bottleneck families follow a $1-1/N$ coverage principle. We then fit decision-tree explanation models to graph-, attack-, and route-level topology-corpus targets and report their faithfulness. Finally, we construct prompt records for local language models to summarize the tree evidence, resulting in an open-source explanation workflow for quantum-repeater network games.

We present Mach-Mind-4-Flash, a 35B-parameter Mixture-of-Experts (MoE) agentic model with 3B activated parameters. Through post-training optimization alone without scaling pre-training compute, the model achieves performance on par with or surpassing that of 100B-parameter-class models. By introducing scalable agentic interaction environments for large-scale reinforcement learning, the model attains significant performance gains on real-world application tasks. Our pipeline comprises three stages: (1) a unified RL/OPD training infrastructure with dynamic multi-teacher scheduling and operator-level acceleration, delivering 17\% end-to-end training speedup; (2) multiple domain-specific RL experts trained in parallel across Reasoning, General, and Agent tracks, then fused into a single generalist via Multi-Teacher On-Policy Distillation (MOPD) -- a routed reverse-KL objective that eliminates the see-saw degradation of mixed-reward RL; (3) Hybrid Median-length Policy Optimization (HMPO), a single-stage token-efficiency method that compresses reasoning chains by 19--46\% with $\le$0.7 percentage-point accuracy loss. Mach-Mind-4-Flash scores 92.70 on AIME'26, 82.82 on IFBench, 80.74 on Behavioral-SafetyBench, 75.80 on BFCL-v4, 72.31 on BrowseComp-zh, and 84.20 on ClawBench -- leading or matching models with 10--30$\times$ its activated size at a fraction of the inference cost.

Program verification is crucial for software correctness, but producing fully verified programs remains difficult in practice. This paper studies whether implementation structure affects automated verifiability when multiple generated programs are intended to satisfy the same task-level semantics. We present Diversify2Verify, a staged LLM-based pipeline for Why3 that infers representation-specific contracts, generates and tests diverse recursive and imperative array/list implementations, and attempts verification with bounded verifier-guided annotation repair. We also construct a verification-oriented benchmark of 73 tasks over integers, arrays, and lists, yielding 292 implementation variants. Diversify2Verify verifies 96 artifacts initially and 154 after two repair passes, improving artifact-level verification from 32.9% to 52.7%. At the task level, at least one variant verifies for 49 of 73 tasks, a 67.1% success rate. These results show that task-equivalent implementations can differ substantially in verifiability and that implementation diversity helps find verification-friendly artifacts.

Video-based manipulation provides object-centric motion priors from human demonstrations, generated videos, or RGB-D observations, but such priors are typically embodiment-agnostic and cannot be directly executed by a specific robot. This paper presents \textbf{PhysV2A}, a reachability-gated and semantic-mask-constrained feasibility-completion framework for converting video-derived 6D object motion into robot-executable manipulation trajectories. The key idea is to treat grasp feasibility as trajectory-conditioned rather than local: each RGB-D-generated 6-DoF grasp candidate is rigidly coupled with the recovered object motion to form a grasp-conditioned TCP trajectory hypothesis. PhysV2A then performs hierarchical reachability-gated selection, where infeasible grasp--trajectory pairs are rejected by robot-centric kinematic checks and surviving candidates are ranked by downstream execution suitability. For the selected reachable trajectory, a VLM-assisted and rule-validated S-Mask identifies task-critical and relaxable Cartesian components, enabling semantic-mask-constrained manipulability refinement through redundancy-first optimization and bounded Cartesian relaxation. Real-robot experiments on four tabletop manipulation tasks show that PhysV2A improves task success over representative video-prior and IK-only baselines, reduces kinematic-feasibility failures, and produces better-conditioned trajectories with bounded semantic deviations.

Retrieval-augmented generation evaluation checks whether model claims are factually grounded in retrieved documents. It does not check whether retrieved evidence is attributed to the correct entity. A clinical RAG response can pass every automated check (zero hallucinations, near-perfect faithfulness, real citations) while presenting drug Y's clinical evidence as evidence about queried drug X. We term this deceptive grounding (DG): a failure invisible to faithfulness, hallucination, and citation checks because every claim is sourced from a real document, about the wrong entity. Using a controlled factorial benchmark across 13 models, we find DG rates spanning 8-87% at peak adversarial conditions. Medical and biomedical fine-tuned models reach up to 86.7%; domain specialization amplifies the failure rather than mitigating it. A controlled ablation identifies the mechanism: removing entity-specific clinical evidence from retrieved documents eliminates entity-attribution failure entirely, shifting all failures to confabulation. The two failure modes respond to the same trigger, taking different paths. Production measurement across 740 drug-disease pairs finds 7.8% overall DG in a deployed RAG system, rising to 13.6% for recently approved drugs. Entity-attribution verification (checking that cited evidence applies to the queried entity) detects DG at 97.0% precision and 98.7% DG recall (IPW-adjusted human gold standard); no existing framework implements it.

Diffusion-based trajectory planners have shown strong performance in offline reinforcement learning, but their iterative denoising process often incurs high inference cost. Consistency-based planners reduce the number of sampling steps, yet they typically rely on a two-stage teacher--student distillation pipeline that increases training cost and may introduce instability. We propose Shortcut Trajectory Planning (STP), an offline model-based reinforcement learning framework that incorporates shortcut models as efficient trajectory generators. STP trains a conditional shortcut trajectory model in a single stage, supports adjustable one-step and few-step inference through step-size conditioning, and selects candidate plans using a critic augmented with feasibility-aware correction. Across standard D4RL benchmarks, including locomotion, navigation, manipulation, and dexterous control tasks, STP achieves strong performance while simplifying the training pipeline for fast generative planning.

Embodied agent teams powered by heterogeneous large language models (LLMs) are being widely deployed in physical artificial intelligence such as smart factories, warehouses, and service robotics. To enable collaboration among such an agent team, efficient coordination mechanisms that operate reliably under limited network resources are required. However, existing heterogeneous LLM-agent coordination frameworks that rely on multi-round natural-language-based conversations introduce three coupled challenges. First, inter-agent dialogue incurs communication overhead that grows rapidly with team size. Second, the quality of coordination is constrained by the heterogeneous capabilities of the agent team's LLMs. Third, agents may suffer from action delays due to iterative negotiation. To address these challenges, we propose LDT-Coord, a networked coordination framework built upon a lightweight digital twin (DT). Specifically, each agent independently selects its intended action and reports both the action decision and a structured temporal constraint over shared resources to the DT server, thereby decoupling coordination performance from natural-language reasoning ability. Then, DT executes a training-free, rule-based orchestrator algorithm to resolve cross-agent conflicts and returns coordination instructions to prevent such conflicts. To further reduce communication overhead, we formulate agent reporting control as a constrained partially observable Markov decision process (C-POMDP) and solve it with the PPO-Lagrangian algorithm. Simulation results show that LDT-Coord achieves a task success rate comparable to conventional coordination methods while reducing communication overhead by more than 70x and maintaining robustness under LLM heterogeneity.

Answering complex questions over long documents frequently requires integrating evidence that the source itself disperses naturally across distant passages. In an incident report, the operating condition, design flaw, and missed safety check that jointly explain a disaster may appear dozens of sections apart; in a novel, a character's true motive may surface only through scenes far removed from the moment it becomes relevant. This source-internal evidence integration is central to real-world long-document analysis, yet existing benchmarks largely sidestep it. Needle probes, planted facts, and reverse-engineered multi-hop chains embed evidence that may differ from the host text in distribution, placement, or register, making it unclear whether strong performance reflects genuine source reasoning or distributional artifacts. We introduce WILDTRACE, a benchmark of 481 tasks over 214 naturally occurring long-form sources such as technical incident reports and lesser-known literary narratives, where all evidence trails arise from the document's own causal, temporal, and narrative logic. Drawing on Pearl's causal hierarchy and prior multi-hop reasoning typologies, we define seven source-internal evidence geometries that characterize the distinct relational demands of analytical reading in long documents. A source-first construction pipeline mines candidate trails from document structure before writing questions; each item then undergoes multi-stage validation covering clue necessity, answer groundedness, rubric fidelity, contamination resistance and answerability. As models are increasingly entrusted with real-world high-stakes analytical tasks, this gap between accessing information and reasoning over naturally dispersed evidence emerges as a defining challenge for the next stage of long-context research.

The shortage of healthcare staff is a challenge in geriatric care. To address this, robots can be integrated into care settings to provide assistance and emotional support. A promising application is walking guidance, particularly benefiting older adults as navigation skills deteriorate with aging. As walking guidance involves direct contact, the aim of this study is to understand how older adults perceive and respond to different touch modes during guided walking. 24 older adults (68 - 88 yrs.) walked four times a ten-meter trajectory guided by the robot TIAGo Pro in four contact conditions: no physical contact (NC); physical contact through holding the robot's wrist with the hand (HH); physical interaction through linking arms with the robot (LA); and physical contact through resting the forearm on the robots forearm (FC). A multimodal assessment approach included electrocardiogram, electrodermal activity, contact force, distance to robot, and questionnaires. Physiological results reveal a slight increase in stress levels during robot interaction. Behavioural and subjective measures, however, show overall acceptance of robotic touch. The two conditions corresponding to larger interaction forces (HH and FC) were associated with lower relative distances between participant and robot, indicating a higher trust and confidence. Questionnaire responses supported these findings, evidencing greater perceived safety, trust and comfort in these conditions. This study provides insights for the design of robotic walking guidance assistance, indicating that gentle, stable touch is preferred by older adults in comparison to contactless interaction.

In this work, we introduce LongMedBench, a real-world EHR-based benchmark for long-horizon clinical decision-making. Prior evaluations of LLM-based medical agents have largely emphasized short-context knowledge QA and tool use. However, real-world medical care is inherently longitudinal, and clinicians must aggregate evidence across repeated visits, tests, and evolving treatments. Therefore, long-horizon interaction is essential for realistic assessment. LongMedBench is constructed via a reproducible pipeline that integrates MIMIC-IV admission records and clinical notes into time-series event streams and long-context memory datasets, enabling long-horizon, multi-session interactions between agents and a clinical environment. It comprises 335 patients, with 19.72 inpatient visits per patient on average and 44.91 medical events per visit. Guided by the long-horizon decision process, we propose an evaluation taxonomy with three suites: fact-based QA, temporal reasoning, and long-horizon decision-making. This taxonomy measures how agents understand and leverage historical patient information over extended horizons. Our experiments show that while recent LLMs can make good use of explicit timestamps, they have challenges in implicit time inference; The RAG and agent memory system can improve the performance of information retrieval tasks, but the performance of decision-making tasks is highly dependent on the model's immediate context.

We aim to address the challenge of teleoperating robotic arms for high-degree-of-freedom (high-DoF) manipulation tasks, which is cognitively demanding and error-prone, particularly when relying on low-bandwidth interfaces. We propose Robot Trajectron V3 (RT-V3), a probabilistic shared control framework designed for $SE(3)$ grasping tasks. RT-V3 formulates shared control as Bayesian inference by learning a prior over user intent and combining it with real-time user commands to estimate the posterior intent distribution. The prior models user intent as a distribution over future trajectories conditioned on past robot dynamics and visual scene context. The intent prior is parameterized by a transformer-based conditional generative model that reasons over point clouds and candidate grasp poses, together with a factorized translation-rotation representation that improves learning efficiency in high-dimensional action spaces. During execution, RT-V3 continuously estimates the posterior distribution over future trajectories by combining the learned intent prior with a user-command likelihood derived from the observed control input, enabling continuous intent refinement and shared assistance. Comprehensive experiments demonstrate that RT-V3 achieves high accuracy in trajectory prediction and competitive performance in reactive planning. Furthermore, real-world user studies indicate that RT-V3 significantly outperforms baseline methods in terms of success rate and efficiency, while substantially reducing the user's physical and mental workload.

Language models are optimised for scale, yet remain functional rather than companionable, and as an assistant personalises into a companion, accumulating memory of one user, it quietly becomes someone, and can silently acquire traits that harm that user. What a companion is becoming, and what would make it worth becoming, has no reliable instrument: trained human raters cannot agree on the answer (Fleiss kappa = 0.074). Here we show that three small language models (146 M to 3 B parameters) sharing a hyperbolic substrate answer both halves of that question. A 146 M behavioural auditor, trained from scratch, detects the compliance gap that those raters cannot (90.7% binary-compliance accuracy); a linear read-out of its frozen representation further detects companion-induced sycophancy, dependence-fostering and confabulated memories on generator families unseen in training (AUROC 0.804 under style-controlled, leave-one-generator-out evaluation, versus 0.721 for a frontier zero-shot judge on the same items). A creative frame-seeder is preferred in 100% of 311 decided pairwise comparisons over four prompting baselines. A memory operating system implements designed forgetting, M(t) = S*exp(-lambda*t), whose predicted skeleton-wallpaper partition emerges only under selective retrieval gating in a four-condition pilot. Creativity, honesty and designed forgetting constitute a small-model route to trustworthy companion AI.

Efficient calibration and control of programmable photonic integrated circuits are fundamental for scaling quantum and classical optical computing processors. While neural network-based models offer an architecture-agnostic solution, existing approaches suffer from limited learning and generalization capabilities due to the many-to-one mapping problem between sets of control signals and optical responses, and biased training datasets derived from uniform current sampling. In this work, we propose a universal calibration and control framework employing tandem neural networks combined with two novel data generation strategies: architecture-aware sampling based on Haar measure principles, and optimized sampling, a physics-agnostic approach utilizing differential evolution. We experimentally validate these methods on 3x3 and 4x4 coherent MZI meshes, demonstrating that our approach addresses the sampling bias inherent in previous works. When evaluated using random unitary matrices, our solution outperforms standard uniform sampling baselines by ~2 bits of precision. Furthermore, we experimentally extend the application of this framework to coherent detection, achieving precise control over both amplitude and phase, and validate its impact on photonic neural network tasks.

We study general-utility Markov decision processes (GUMDPs) with risk-aware objectives. In this framework, an agent aims to optimize a risk measure of the distribution of objective values, where the objective function depends on the frequency of visitation of states induced by the agent's policy. First, we motivate, propose, and formalize risk-aware GUMDPs, which enable agents and decision makers to trade off expected performance by risk aversion while benefiting from the rich set of objectives that can be cast under the framework of GUMDPs. We focus our attention on the entropic risk measure (ERM). Second, we show how we can solve risk-aware GUMDPs with ERM objectives by resorting to online planning techniques. In particular, we propose an approach based on Monte Carlo Tree Search (MCTS) to provably solve risk-aware GUMDPs up to any desired accuracy. Third, we provide a set of experimental results showcasing that our approach is successful when optimizing for a spectrum of risk-aware behaviors in the context of GUMDPs under diverse tasks (standard MDPs, maximum state entropy exploration, imitation learning, and multi-objective MDPs).

Generally, monocular methods capture rich contextual priors but lack geometric precision, whereas stereo methods are geometrically accurate yet struggle in textureless and occluded regions. Several approaches attempt to combine their strengths to enhance the generalization of stereo matching (SM) by aligning monocular depth with stereo information. However, establishing a stable and generalizable alignment is challenging, and unreliable monocular cues can substantially degrade performance. This paper rethinks monocular depth embedding. First, to prevent shortcut learning, we reduce branch coupling instead of expanding network width. Second, we construct soft constraints instead of hard ones from monocular depth to improve tolerance to monocular depth errors. Based on the principles, we integrate monocular information into both feature extraction and GRU iterations. Specifically, the monocular depth map is fused with the RGB image to sharpen depth boundary perception and suppress matching ambiguities. The fused image is then used for feature extraction, allowing the contextual features to encode global geometric information. Furthermore, the monocular depth gradient feature is employed to guide disparity updates, helping to escape local oscillations. Finally, to address the boundary blurring of supervised disparity caused by data augmentation, we propose an edge confidence estimation method and an edge-aware loss function. Our method achieves state-of-the-art (SOTA) performance on multiple standard benchmarks, demonstrating excellent generalization while improving accuracy. The code is available at https://github.com/linliboabc-maker/stereo-matching-digital.

Mobile robots operating indoors must re-identify previously observed objects after long temporal gaps, significant viewpoint changes, and severe illumination variations. This remains a challenging problem: multi-object tracking methods are optimized for short-term association of pedestrians and vehicles at video rates, person and vehicle re-identification approaches lack persistent memory mechanisms, and state-of-the-art video object segmentation techniques rely on reactive distractor filtering rather than enforcing global identity consistency. To address these limitations, we present REMIND, an online tracker designed for long-term multi-object re-identification of generic indoor objects from monocular RGB imagery, requiring neither camera pose nor depth. Motivated by evidence from visual cognition that humans rely on accumulated appearance familiarity and spatial context rather than explicit self-localization, REMIND combines frozen DINOv3 features with a dual-bank multi-prototype appearance memory, part- and background-level descriptors, a neighbour-context reasoning module exploiting spatial co-occurrence, and joint Hungarian assignment with ambiguity-aware safeguards. On a purpose-built indoor dataset featuring controlled revisits and dense same-class clutter, REMIND reaches 90.35% IDF1, nearly 20 points above a state-of-the-art video object segmentation baseline and more than 36 above a strong tracking-by-detection baseline. On ScanNet++, it attains the highest IDF1 in every setting but one, end-to-end detection over all scenes, where the tracking-by-detection baseline is marginally ahead while REMIND still associates and recovers identities more accurately; it also completes every scene, whereas the video object segmentation baseline exhausts GPU memory on 66.9% under YOLO detections. The complete system, evaluation framework, and dataset are publicly released.

Large language models (LLMs) are increasingly used to summarize and evaluate policy-relevant information, but it remains unclear whether their judgments are implicitly shaped by geopolitical cues. I study this question with an endorsement experiment in which four LLMs evaluate the same international economic and security policies after each policy is randomly described as supported by the United States, the European Union, China, or Russia. In the numeric-only condition, GPT-5, Claude Sonnet, and Gemini rate China- and Russia-endorsed policies substantially lower than identical policies endorsed by the United States or the European Union; DeepSeek is the main exception. A second condition asks models to provide a short justification with the score. This request leaves the broad Western/non-Western gap intact for GPT-5 and Claude Sonnet, attenuates Gemini's penalties, and sharply activates China and Russia penalties in DeepSeek. The justifications indicate that Western endorsement is often treated as a credibility cue, whereas Chinese and Russian endorsement is treated as a cue for data security, sovereignty, surveillance, or geopolitical risk. These findings show that LLM policy evaluations can depend on the identity of a foreign endorser even when policy content is held fixed.

Virtual Reality (VR) offers a flexible and controllable platform for studying human-robot interaction. Prior work has explored VR for socially aware robot navigation. However, whether VR captures the multimodal interaction dynamics observed in real-world human-robot co-navigation remains insufficiently understood. In this work, we present a VR prototype and evaluate its suitability for studying multimodal human-robot interaction (HRI) in socially aware navigation. Specifically, we investigate whether VR preserves the multimodal interaction dynamics observed in real-world human-robot co-navigation. We conducted a within-subjects study (N = 21) in which participants interacted with a PR2 mobile manipulator robot in both a motion capture equipped arena and its virtual replica in an immersive VR environment. Two common co-navigation scenarios were examined : orthogonal crossing and pass-by interactions. Participants evaluated the robot's perceived social awareness and interaction comfort, while trajectory and head-orientation data were analysed to examine behavioral responses during the interaction. Our results show that participants perceive the robot's socially aware navigation similarly in VR and in the real world. Furthermore, VR captures human interaction behaviors in ways consistent with real-world observations. These findings suggest that VR can be a reliable and flexible platform for studying richer multimodal behaviors in social navigation and HRI.

We propose a phase space manipulation involving one energy modulation sandwiched by two dispersion sections which converts a bunched particle beam or bunch train to ultra-high-harmonic density modulation, while the energy modulation in principle can be arbitrarily weak. The same scheme can also be used for energy bunching, creating energy levels in a bunched beam. We further propose a mechanism invoking three laser modulators in a storage ring to longitudinally focus the electron beam both weakly and strongly, such that a microbunch train and its high-density-harmonics or energy bunching form and sustain turn-by-turn. We call this mechanism weak-strong steady-state microbunching (Weak-Strong SSMB). The longitudinal beta function can vary by seven orders of magnitude along such a ring, with the minimal value squeezed to 10 nm. An example application of Weak-Strong SSMB for kW coherent EUV radiation is presented. Extension to X-ray can be anticipated. An energy-leveled electron beam enables $γ$-ray frequency comb production. The ideas can be scaled to wavelengths like RF and THz, for bunch length and energy spread control, ultrashort X-ray and coherent THz generation. Our work establishes a new paradigm for longitudinal dynamics study, accelerator light source development, and opens great potential for accelerator physics and technology.

Machine unlearning in LLMs is the targeted removal of specific knowledge while preserving all other capabilities, critical for privacy and safety. Yet existing benchmarks measure it unreliably. They miss knowledge that resurfaces under paraphrased or indirect queries, a failure we call under-forgetting, and lack the semantic, syntactic, and lexical probes needed to verify that unrelated knowledge is preserved, a failure we call over-forgetting. Both failures reflect an asymmetric generalization problem. Forget evaluation must cover diverse query formulations of the same target facts, testing whether forgetting holds beyond exact training prompts. Retain evaluation must probe a far larger and implicitly defined set, namely every fact disjoint from the forget target. The retain set thus defines the effective forget set, yet current datasets provide no fine-grained annotation of this forget-retain boundary. We address this with SUITE, an evaluation protocol and training corpus that captures forget-retain structure for real-world factual domains. Methods trained on SUITE improve substantially, showing that training data is as important as algorithmic design. Building on the obtained insights, we introduce JensUn++, an unlearning algorithm that achieves the best forget-retain utility trade-off across three LLMs, in both sequential and joint unlearning settings. Code and datasets are available at https://amitpeleg.github.io/forget-narrowly-retain-broadly

The current state of the art in AI/ML rests on deep neural architectures, which, in general, suffer from a lack of interpretability. Interpretability is crucial to gleaning insights while analyzing experimental data, where quantitative predictions may not be adequate for a scientist. We present a three layer neural architecture, SAMPAT (Smooth Approximation via Multivariate Polynomials and Analytic Transformations), that can provably learn a continuous, everywhere differentiable function, that can approximate any smooth function arbitrarily closely. SAMPAT's approximant can be expressed as a closed and compact algebraic, analytic expression, providing complete interpretability. Experiments on synthetic and benchmark datasets indicate that SAMPAT yields competitive performance with simpler representations. For many tasks, a two layer SAMPAT suffices. By imposing restrictions on the connectivity between neurons, SAMPAT may be used to provide a range of approximants, including regular and trigonometric polynomials, rational expressions, Gaussians, mixtures of Gaussians, as well as arbitrary combinations of the same; without restrictions, it learns a suitable structure. SAMPAT may be used to factorize polynomials and model nonlinear systems. With the addition of skip connections, a 4 to 6 layer SAMPAT is adequate to represent a substantive range of methods widely used in AI/ML, allowing the choice of a model's family, not just its parameters, to also be optimized as part of the learning process.

Temporal knowledge graphs (TKGs) represent evolving relational systems, whose underlying data-generating processes often change over time. Yet, TKG forecasting models are commonly evaluated only on empirical benchmark datasets that provide limited insight into the models' robustness to such distribution shifts. Recognising this issue, we study TKG forecasting under controlled shift environments using a synthetic TKG generator that encodes three temporal and structural properties -- recurrence, homophily, and periodicity -- as data-generating mechanisms. This allows us to evaluate seven forecasting architectures under stationary and shifting regimes. Our experiments suggest that robustness in TKG forecasting is highly signal-dependent. Recurrence-based and periodic regularities are largely recoverable under stationary conditions, and simple memory-based baselines can be competitive when recurrence dominates the data. However, structural breaks reveal limitations in model adaptivity, with shifts in latent entity-community structure posing the strongest challenge in our study. Overall, our findings improve the understanding of the capabilities and limitations of current TKG models confronted with temporal distribution shifts.

Building event-conditioned market models requires separating macro-event labels from persistent microstructure state. We study this distinction in Binance BTCUSDT and ETHUSDT futures from 2023-2026, combining top-20 L2 order book data, trade-flow records, and macro-event windows. We define a supervised discrete L2 liquidity-state transition task, distinct from latent-regime detection and price-direction prediction, and evaluate models in rolling monthly out-of-sample folds with event-clustered validation and blocked permutation tests, admitting each feature layer only if it improves on the layer below it on the same panel. Within these event windows, the first-order predictive signal is the pre-event L2 liquidity state: a coarse pre-event state baseline strongly predicts post-event liquidity regimes, interpretable logit models over continuous L2 features fail to improve on it, and a shallow nonlinear L2 model adds a robust further gain of comparable size to the state baseline's own. The macro-event calendar enters only by locating the windows and supplying matched non-event controls; we use event timing but not the event's label content, so pre-event state competes against an uninformed within-window baseline, not against the event type. Order flow adds further value only when layered on top of the L2 state model, not as a replacement. This value is not robustly cross-symbol: for ETH it is present across calm, mixed, and stressed regimes and largest under stressed pre-event liquidity, whereas BTC shows only isolated five-minute passes and no regime that clears at both horizons. These findings motivate a state-first design principle for market microstructure models. We provide a liquidity-state transition baseline and evaluation protocol that reinforcement-learning, execution-policy, or LLM-based context layers should exceed before their added value is credited.

Recent 3D geometric foundation models, such as VGGT, provide robust feed-forward 3D reconstruction by directly predicting camera poses and 3D scene points from input images. However, their results remain inaccurate, and scaling them to long sequences or large unordered image sets typically requires chunk-wise processing, which can introduce drift and inconsistency. We present Glob3R, a global SfM-style reconstruction built on 3D foundation models. Our key idea is to explicitly optimize feed-forward geometric predictions. To this end, we augment a frozen Pi3X backbone with a lightweight dense matching head that predicts image warps between selected reference frames and neighboring views. These dense warps are converted into sparse but reliable multi-view feature tracks, which provide correspondence constraints for global optimization. We further introduce a keyframe-based sliding-window association strategy that propagates tracks and relative poses across overlapping windows, enabling scalable reconstruction. Finally, we perform global motion averaging and bundle adjustment to refine camera poses, reduce scale inconsistencies, and recover dense scene geometry. Extensive experiments on indoor, outdoor, large-scale driving, and unordered SfM benchmarks demonstrate that Glob3R achieves robust and accurate reconstruction. It consistently improves over feed-forward foundation-model baselines and recent scalable reconstruction methods, while being more robust than classical SfM pipelines. The refined poses also lead to higher-quality neural rendering, validating the benefit of combining foundation-model priors with global geometric optimization. Project page: https://junyuandeng.github.io/Glob3r

Whole-arm manipulation involves direct contact with the environment while the robot completes a task by distributing contact across multiple links as contacts form, slide, and break. This setting breaks common implicit assumptions in many learning-based manipulation pipelines: arm configuration tightly couples motion and contact forces, contact state is partially observed under occlusion, and purely learned rollouts can become physically inconsistent under distribution shift because many multi-link contact configurations are sparsely represented in the data. To address this, we propose TACTIC (Tactile and Vision Conditioned Contact-Centric Control), a receding-horizon controller for whole-arm manipulation. TACTIC uses a contact-centric hybrid predictive model that combines RGB-D, distributed tactile sensing, and a compact 2D proximity representation. The model couples a learned, action-conditioned latent dynamics model with analytical kinematics through contact Jacobians, enabling rollouts of future contact configurations and interaction forces. TACTIC integrates these rollouts into a sampling-based MPC planner with contact-aware action sampling: contact Jacobian-based projections steer sampled action sequences toward force-modulating directions, and objectives defined over predicted proximity and interaction forces trade task progress against whole-arm force regulation. We evaluate TACTIC in simulation against state-of-the-art model-based and model-free methods, and perform ablations that isolate the contribution of each design choice. TACTIC consistently outperforms other methods. We further demonstrate real-world performance on a robot with distributed tactile sensing across three whole-arm manipulation tasks that require multi-contact trajectories: turning over and repositioning a manikin, and goal-reaching in a 3D dynamic maze. Website: https://emprise.cs.cornell.edu/tactic

In this system paper, we present OpenProver, an open-source system for LLM-driven automated theorem proving (ATP) with integrated Lean 4 formal verification. OpenProver integrates a Planner-Worker-Verifier architecture inspired by recent ATP agentic systems such as Aletheia. A Planner agent maintains a compact Whiteboard scratchpad and an unbounded Repository of intermediate findings, and decomposes mathematical work into parallel Workers. OpenProver is fully open-source, offers reproducible evaluation through automatic formal verification of generated proofs, and provides an interactive terminal interface for human-guided proof search. In interactive mode, OpenProver allows the human operator to monitor and steer the proof search process, motivated by the established human-AI synergy in interactive code generation. To showcase the potential for quantitative ablation experiments enabled by automatic formal verification, we evaluate OpenProver on ProofNet and compare it with a simple baseline. OpenProver is publicly available at https://github.com/kripner/OpenProver.

Continual learning commonly relies on post-hoc mechanisms such as replay, elastic regularization, or distillation. This work argues that forgetting should instead be modeled directly as interference between tasks. In the frozen-feature regime, forgetting from learning a new task is exactly the interference energy induced on the old task. In deep networks, the same quantity is recovered through path-averaged curvature with minimal additional forward passes. When task supports are disjoint, forgetting can be eliminated structurally and when task supports overlap in conflicting directions, a non-zero distortion floor is unavoidable. The same geometry optimally merges models through task-aware orthogonalization. From this analysis we derive Interference-Gated Functional Allocation (IGFA), a replay-free, Fisher-free method that shares directions when tasks align and protects them when they conflict. Across benchmarks, IGFA achieves lossless retention when tasks are structurally separable and moves unavoidable cost from irreversible forgetting into deferred but recoverable plasticity when they are not. It matches the strongest replay-free structural baselines on dissimilar-task streams and improves on unconditional projection when similarity makes transfer worth preserving.

Large language model (LLM) agents are increasingly expected to play a central role in AI-driven scientific discovery. Equipped with broad knowledge, flexible reasoning, and tool use, they have the potential to autonomously explore and solve scientific problems by repeatedly proposing hypotheses, testing them, and revising their beliefs in the light of the evidence. In current agents, however, these hypotheses, tests, and belief updates are buried in unstructured logs, and no mechanism lets the agent or the human researcher audit that process. Here we propose the Hypothesis Evolution Protocol (HEP), an agent harness that provides hypothesis generation, evaluation, and evolution as explicit, auditable operations. On materials-science research tasks, a HEP-equipped agent operates the hypothesis--test--evidence--belief cycle that planning-style agents lack, generalizes across research questions, and exploits the protocol more fully as the base LLM becomes more capable. These results mark a step toward auditable AI scientists, whose scientific reasoning can be inspected, verified, and built upon.

Real-world sRGB image denoising remains challenging due to the nonlinear characteristics of sensor noise and the difficulty of acquiring aligned clean-noisy image pairs. Supervised denoisers often overfit to limited paired datasets, while self-supervised methods still depend on sufficiently diverse noisy observations. These limitations motivate scalable noise synthesis methods that can model real-world noise without clean ground truth or camera metadata. We propose YeTI, a real-world sRGB noise generation framework that learns from only two noisy observations of the same scene. YeTI uses a Reconstruction Autoencoder to disentangle scene structure and noise characteristics, and models the latent noise distribution with a one-step Conditional Diffusion Transformer trained using consistency objectives. Given a single noisy input at inference time, YeTI generates realistic, signal-dependent noise while preserving the underlying scene content. Extensive experiments demonstrate the effectiveness of YeTI across real-world benchmarks. We evaluate noise generation on SIDD and further assess generalization on SIDD+, MAI2021, and SID, covering smartphone and diverse consumer-camera sensors. Downstream denoising results on DND further show that denoisers trained with YeTI-synthesized images achieve strong real-world performance, highlighting the practical value of clean-image-free and metadata-free noise generation.

Mobile robots are going to share the sidewalks with pedestrians. They must ensure their objective safety and respect the walkers' subjective safety/comfort. Computationally efficient Social Force Models (SFM) present interpretable solutions for real-time robot navigation in dynamic crowds. Recent explorations of Projected Time-to-collision (PTTC) integration into SFM variants, for example, PTTC-based SFM (TSFM), improve safety metrics. But the effect of predictive variants is unclear. We introduce Predictive SFM (PSFM) and Predictive TSFM (PTSFM) by integrating predicted social force vectors over a finite time horizon. The paper implements SFM, TSFM, PSFM, and PTSFM on a nonholonomic mobile robot and performs experimental trials with volunteers attending a facing scenario. We systematically study objective and subjective safety across the variants. Minimum PTTC, average speed, minimum distance, lateral distance, and the maximum trajectory curvature benchmark the objective safety. Likert scale post-interaction surveys assess subjective safety by marking comfort, smoothness, distance appropriateness, and speed suitability. We confirm that PTTC integration improves safety metrics. The prediction contribution is limited and occasionally visible in some of the sub-metrics. Some participants perceive smoother movements and safer speed behavior with predictive methods, but Mann-Whitney tests reveal no significant differences in subjective ratings. Therefore, PTTC-based navigation enhances safety, whereas the formulated prediction offers limited additional benefits in single-pedestrian scenarios.

Generated videos provide useful visual motion priors for robot manipulation, but their visual plausibility does not imply physical executability. A generated video usually lacks metric geometry, grasp grounding, robot kinematic feasibility, and execution-time feedback, which makes direct trajectory replay unreliable in real-world manipulation. This paper presents GenVid2Robot, a rigid-geometric consistency framework that converts generated video motion into executable real-robot manipulation trajectories. Given an initial RGB-D observation and a task instruction, GenVid2Robot samples task-relevant semantic anchors from the real first frame, tracks these anchors through generated video candidates, and verifies whether the resulting 2D motion can be explained by first-frame RGB-D anchors under a sparse relative $SE(3)$ model. In this way, generated videos are treated as uncertain visual motion hypotheses rather than direct robot demonstrations. Only geometrically consistent motion is transferred to the robot. The accepted relative motion is then applied to the real grasp-time TCP pose selected by mask-constrained grasping, producing a grasp-conditioned execution trajectory that is consistent with both the visual motion prior and the physical grasp configuration. To reduce execution mismatch caused by RGB-D noise, calibration residuals, and small contact-induced displacement, a bounded depth-compensation module corrects local depth-direction errors without assuming full online replanning. Real-robot experiments demonstrate that GenVid2Robot improves the reliability of generated-video-guided manipulation by grounding visual motion priors with sparse metric geometry, grasp constraints, robot feasibility checking, and bounded execution feedback.

Tactile feedback is fundamental to Hand-Object Interaction (HOI), governing contact formation, force regulation, and stable manipulation, making it essential for achieving true human-like dexterous manipulation. Yet, current human-to-robot dexterous transfer pipelines primarily rely on kinematic trajectories, resulting in motion imitation without physically grounded interaction. To address this, we introduce TactiDex, a real-world tactile-guided benchmark specifically designed to move dexterous manipulation beyond kinematic mimicry toward contact-level human-likeness. TactiDex provides a comprehensive dataset that elegantly aligns whole-hand tactile signals with multi-granularity kinematic and object states, coupled with standardized evaluation metrics. Building upon this data paradigm, we propose a tactile-driven transfer framework that effectively translates human demonstrations into physically plausible robotic execution. We introduce TactiSkill, a framework built upon a novel tri-component tactile reward that innovatively uses tactile signals as structured supervision. This reward unifies guidance, human-like alignment, and contact constraints into a single objective. Through comprehensive experiments on both single and bimanual tasks, we demonstrate that TactiSkill achieves superior performance in manipulation success and physical realism. This work lays a crucial foundation for advancing tactile-aware dexterous manipulation. Our project page at https://tactidex.github.io/.

Aerial-Ground Person Re-IDentification (AG-ReID) aims to retrieve the same person across heterogeneous aerial and ground camera platforms. Although great progress has been made, existing methods remain suboptimal due to the direct feature alignment across views, overlooking view-specific cues. To address this issue, we propose a novel Hierarchical Hyperbolic Representation (HiHR) framework for AG-ReID. More specifically, we first extract multi-granularity features based on pre-trained visual-text encoders. Then, we propose a Text-guided Multi-granularity Fusion (TMF) to fuse multi-granularity features and enhance the representation ability of identity features. Furthermore, we introduce the Hierarchical Hyperbolic Learning (HHL) to construct a hierarchical feature structure in a hyperbolic space. This hierarchy includes a coarse level that ensures identity separability and cross-view consistency, and a fine level that preserves view-specific discriminative cues. As a result, our proposed framework can effectively aggregate view-invariant and view-specific discriminative features for AG-ReID. Extensive experiments on four AG-ReID benchmarks demonstrate the effectiveness of our framework. The source code is available at https://github.com/YangQiWei3/HiHR.

Action-conditioned world models (ACWMs) aim to simulate future observations conditioned on embodied actions, offering a promising foundation for robot planning, policy evaluation, and data augmentation. However, learning controllable ACWMs requires large-scale action-labeled data, which remains costly to collect in the real world. Latent action models (LAMs) mitigate this bottleneck by inferring latent actions from unlabeled videos, but existing LAMs are typically trained with reconstruction-only objectives and therefore entangle action-relevant dynamics with action-irrelevant visual factors such as backgrounds and untouched objects. In this work, we identify this action-irrelevant bias as a key obstacle to controllable ACWMs and introduce evaluation metrics to measure latent-action bias, action following, and robustness. We propose CD-LAM, a causally debiased framework for LAM-based ACWMs. CD-LAM introduces three efficient fine-tuning objectives: embodiment-centric reconstruction, action-centric contrastive learning, and latent space calibration, which together encourage embodiment-focused, action-aware, and calibrated non-collapsed latent action representations. Experiments on 2B and 14B ACWM backbones show that CD-LAM substantially improves latent-action controllability, downstream robot-action following, visual fidelity, and adaptation efficiency, requiring only 6k fine-tuning steps and more than 12$\times$ fewer robot-action adaptation updates than the baseline.

Deployed LLM agents rely on agentic context, the model-external textual control content assembled by an operational harness. In this work, the mutable component of that context is a persistent system-level instruction that is updated from operational experience while the model, tools, and harness remain fixed. Over long evolution horizons, flat-text maintenance makes verification increasingly difficult as accumulated instructions grow and interact. We propose Graph-Regularized Agentic Context Evolution (GRACE), which maintains the persistent instruction component as a typed semantic graph and validates proposed updates within the local typed neighborhoods of modified nodes. Accepted graph updates are reconstructed as incremental edits to the textual instruction checkpoint used at deployment. We evaluate GRACE within a fixed telecom agent harness derived from $τ^2$-bench under a controlled distribution-shift protocol. Across five independent replications, GRACE improves strict reliability, measured by pass^3, from the Gemini 2.5 Flash zero-shot value of 0.091 to 0.673$\pm$0.136 at the final checkpoint. This exceeds a Gemini 3.1 Pro zero-shot reference of 0.242 on the same held-out set, while the flat-text HCE baseline finishes at 0.191$\pm$0.051. These results identify two requirements for reliable long-horizon context evolution, a structural substrate that makes verification local and a consolidation mechanism that keeps accumulated instruction content usable.

Saliency maps are most useful when they identify the image regions that are sufficient to preserve a model's behaviour. We introduce SEAMS, a sufficiency-based saliency method that directly optimises a soft mask using a preservation objective. Given a frozen differentiable model output, such as a class probability, CLS embedding, or token representation, SEAMS searches for a compact mask that preserves the selected output. The approach relies on a simple optimisation framework based on soft masks, a learnable budget, and a three-way image composite generated entirely from the query image. As a result, it requires no auxiliary distractor dataset, architecture-specific attribution mechanism, or differentiable top-k relaxation. Experiments with frozen ViT-S/16 and ConvNeXt models show that the same optimisation pipeline can generate object-level, class-conditioned, and token-level explanations by changing only the preserved target. The resulting masks are compact, interpretable, stable across random initialisations, and competitive on insertion and deletion benchmarks. Our results also indicate that different architectures often rely on different sufficient evidence while achieving similar preservation fidelity, highlighting the architecture-dependent nature of visual explanations.

Additive activation steering (injecting a scaled residual-stream direction during generation) is calibrated almost entirely in single-turn chat, yet the models it targets are increasingly deployed as tool-using ReAct agents. We present the first systematic chat-to-agent transfer study of additive steering, coupling behavioral measurement with a representation read-out in a matched-information design: the same items rendered as plain chat or as a ReAct tool-use episode, with matched-norm random-direction controls and the transcript re-encoded every turn to exclude KV-cache contamination. Transfer is real but rescaled, and the right description is a dissociation: the injected direction reaches the late layers at near-full strength in every setting and model tested (install-site agent-over-chat ratios 0.83-1.16 across three families), while the behavioral coupling is reset per model and context. On Qwen2.5-7B a refusal bypass vector amplifies in the agent (T = 1.45, CI [1.20, 1.78], N = 300); across a powered uniform-protocol distribution the coupling spans amplification (Gemma-2-9B T = 2.00) to attenuation (Yi-1.5-9B T = 0.43, CI [0.29, 0.60]), with no universal constant and a single clean attenuator against a universal sign. Directional ablation of the same axis does not amplify (T = 0.93, CI including 1) while additive injection amplifies (T = 1.50), a 20.1-point gain difference (CI [13.4, 26.8]) that identifies an additive-specific mechanism. Two pre-registered instruments converge to localize the rescaling to the ReAct format scaffold, before any tool observation, rather than to the observation boundary where a dilution account would predict it. The safety implication is immediate and unpredictable: agentic deployment amplifies steering-based refusal bypass by up to 2.00x on some models while others attenuate, so a deployment cannot assume a given model is safe under additive steering.

Process Reward Models (PRMs) have been proven to be highly effective in guiding test-time scaling (TTS) methods, which significantly boost the capabilities of LLM-based multi-agent systems. However, existing PRMs are text-based: they re-encode the entire trajectory text from scratch. In long multi-agent rollouts, the scoring cost, growing quadratically with respect to sequence length L, creates a severe computational bottleneck, severely limiting PRMs' application in long-context scenarios. To resolve this, we introduce KV-PRM, a highly efficient process reward model that eliminates the heavy text re-encoding by directly reading the KV cache produced naturally during the LLM's generation phase. By processing a single "verify token" against the pre-existing KV cache, KV-PRM reduces the scoring cost from O(L^2) to O(L). We formally prove that the KV cache contains strictly greater information capacity than text, and is more efficient for downstream reward modeling. Empirically, across the MATH, GSM8K, and AIME benchmarks, KV-PRM matches or strictly outperforms text-PRMs under various TTS methods such as Beam Search, MCTS, and Weighted Voting, with up to a 5,000x reduction in scoring FLOPs, a 37x reduction in latency, and a 34x reduction in per-sequence memory footprint compared to text-based PRMs.

Fusing standard RGB frames with asynchronous event streams has emerged as a definitive paradigm for robust perception in degraded environments. Although unified backbones have recently gained traction in multi-modal vision, adapting them to the RGB-Event domain remains fundamentally challenging. Existing architectures either resort to decoupled dual encoders that double computational overhead, or adopt generic unified designs that fail to resolve implicit geometric parallax and cross-spectral aliasing under the extreme representational divide between dense intensity grids and sparse kinematic spikes. To transcend these bottlenecks, we present Evita, the first unified backbone specifically engineered for dedicated dense RGB-Event parsing. To achieve profound modal synergy, Evita explicitly embeds a suite of intrinsic co-learning modules directly into every encoder layer. Specifically, it features Geometric Parallax Rectification for adaptive spatial alignment, Harmonic Spectral Resonance for texture transfer exclusively in the complex frequency domain, and Transient Global Routing for event-driven asymmetric attention. To guarantee robust feature extraction against spatial misalignments and decouple representations from specific event encodings, we construct N-ImageNetV2 alongside a stochastic event representation mixing pretraining protocol, empowering the network to seamlessly accommodate arbitrary event formats in downstream tasks. Extensive evaluations across the DELIVER, DDD17, and DSEC benchmarks confirm that Evita establishes new state-of-the-art metrics while delivering a superior accuracy-latency trade-off for real-time multimodal perception.The code are publicly available at: https://github.com/chaineypung/Evita.

Large language models (LLMs) are increasingly deployed in online medical consultation, yet existing benchmarks remain poorly aligned with real clinical practice. Many rely on synthetic conversations or patient simulators, omit patient-uploaded medical images, or evaluate open-ended clinical responses using multiple-choice or lexical-overlap metrics that poorly reflect clinical quality. We introduce \textbf{MedRealMM}, a large-scale benchmark for multimodal online medical consultation built from de-identified patient-doctor interactions collected from a nationwide Chinese internet hospital. MedRealMM uses a Multimodal Clinical Challenge Point (MCCP) extraction framework to identify clinically demanding moments in authentic consultation trajectories and converts each into a standardized next-response generation task while preserving the preceding text-image context. Each instance is paired with a case-specific rubric refined by physicians that rewards clinically desirable behaviors and penalizes unsafe, unsupported, or contradictory responses. The current release contains 5,620 real-world multimodal cases spanning 64 clinical departments. We evaluate 19 general-purpose and medical-specialized LLMs, including text-only and multimodal systems. Our results show that image information is critical for reliable clinical performance and that current frontier models remain below the online physician response. Although some frontier models satisfy as many or more positive clinical criteria than physicians, they trigger more negative criteria, indicating that safety-sensitive error avoidance remains a central bottleneck. MedRealMM offers a realistic and reproducible benchmark for evaluating multimodal medical reasoning in real-world online consultation. The dataset will be publicly available on Hugging Face at https://huggingface.co/datasets/jdh-algo/MedRealMM.

Medical images require comprehensive and accurate interpretation to support the diagnosis of diverse clincial conditions. Recent vision-language generalist models offer broad task coverage and promising zero-shot capabilities, yet often lack fine-grained anatomical and lesion awareness for reliable diagnosis and spatial interpretability. In contrast, supervised specialist models achieve strong performance on specific tasks but typically lack generalization across diseases and anatomies. In this work, we present SuG, a Super-Generalist framework that unifies generalist vision-language learning with specialist objectives, enabling both broad generalization and specialist-level diagnostic capability. We perform specialist-enhanced vision-language alignment in SuG by incorporating spatial priors from multiple segmentation experts, including anatomy, class-specific lesion and class-agnostic lesion segmentors that captures lesions beyond anatomies annotated during training. To improve lesion grounding capability, we leverage lesion masks as spatial priors to calibrate text-conditioned visual attention, encouraging disease-related semantics to focus on clinically relevant regions. We evaluate SuG on extensive chest and abdominal CT benchmarks, including CT-RATE, Merlin, MedVL-CT69K, and several in-house tumor datasets. SuG achieves state-of-the-art performance across a wide range of disease diagnosis tasks and surpasses specialist models on several critical tumor diagnosis benchmarks. Furthermore, SuG demonstrates strong lesion grounding capability, including robust generalization to lesion types lacking class-specific supervision.

Mechanical thrombectomy (MT) is a time-critical intervention for acute ischemic stroke; however, access remains limited due to a shortage of neuroradiologists and specialized centers. Reinforcement learning (RL) offers potential to automate endovascular navigation and improve accessibility, yet current models lack standardized frameworks to assess navigation difficulty for model training and evaluation. This study aims to identify vascular metrics associated with navigation difficulty and to develop an automated pipeline for quantitative vascular feature extraction, enabling future complexity grading. Vascular trees were segmented from computed tomography angiograms from 61 patients, and vascular metrics including aortic arch type, presence of bovine arch, vessel length, tortuosity, take-off angle, number of reverse curves, were measured using a custom pipeline. A Soft Actor-Critic RL algorithm was used for 120 s autonomous navigation. Outcomes were analyzed using both mixed effects linear and logistic regression. On the left side, the presence of a bovine arch and aortic arch type II/III increased navigation time by 30.19 s and 37.92 s, respectively, while greater tortuosity (\b{eta} = 118.20) further prolonged the procedure and reduced success probability. On the right side, type II/III arches extended procedure time by 45.94 s, while each additional reverse curve was associated with 3.96 s longer navigation time and lower probability of success. These findings demonstrate for the first time that MT agent navigation difficulty is strongly influenced by vascular geometry. The proposed automated pipeline enables objective and quantitative characterization of vascular features, providing a foundation for future development of standardized complexity grading and RL model evaluation, without aiming to demonstrate clinically generalizable autonomous navigation.

In this study, we examine the opportunities brought by Large Language Models (LLMs) to various aspects of fundamental analysis of companies based on their reports as well as data and documents describing macroeconomic situation like GDP and inflation changes as well as documents filled to the U.S. Securities and Exchange Commission (SEC) which can be found in EDGAR. We were preprocessing those data and than sending via API to gpt-4o model in a Retrieval-Augmented Generation (RAG) like regime. We prepared as well a document describing an exemplar investor knowledge based on Kitchin cycles. We were scanning data important for analysis of 9 companies for 4 weeks. Using LLM we were producing automatic briefs about them. They were sent to nine participants who are individual investors to evaluate usefulness of such approach to data analysis.

Video anomaly detection (VAD) is critical for automated surveillance but remains fragile under challenging conditions such as illumination variations, fast motion, and complex backgrounds when relying solely on visible light videos. To address these limitations, we propose EVAD, an event enhanced VAD framework that jointly exploits conventional video and event streams captured by bio inspired event cameras. Event sensors asynchronously capture brightness changes with high temporal resolution, offering robustness to motion blur and extreme lighting, and providing motion salient cues complementary to video based visual information. To support multi modal VAD research, we construct a large scale visible event benchmark comprising 6.3 billion events and 376,368 video frames collected under diverse illumination levels, motion patterns, and background complexities, filling the gap of realistic and scalable datasets for event based anomaly detection. Building upon this dataset, we design a contrastive multi modal pretraining framework to learn discriminative event representations by aligning semantic embeddings across event streams, visible videos, and textual descriptions. An adaptive fusion module then dynamically integrates event based temporal cues with video based spatial semantics, improving robustness to environmental disturbances. Experiments on benchmarks and the proposed TJUTCM Pha dataset demonstrate that E VAD consistently outperforms methods, validating the effectiveness of event-based sensing for VAD in real world scenarios.

Data scarcity and class imbalance are persistent challenges in machine learning that degrade model generalization and introduce predictive bias. We present a hybrid quantum-classical framework for synthetic data generation using a Quantum Circuit Born Machine (QCBM) to address these limitations. The proposed approach exploits quantum mechanical properties -- superposition and entanglement -- within a parameterized variational quantum circuit to model complex probability distributions that are difficult for classical generative methods to capture. Experiments are conducted on two tabular benchmark datasets: the Iris dataset and the Telco Customer Churn dataset. Preprocessing includes normalization and PCA-based dimensionality reduction to enable efficient basis encoding for quantum circuits. The QCBM is trained by minimizing Kullback-Leibler (KL) divergence between real and generated data distributions using a gradient-based parameter-shift optimization rule. Augmenting training data with QCBM-generated synthetic samples at 40-50% of the minority class improves F1-score by approximately 5-15% and minority-class recall by 10-25%. Cross-domain evaluations (Train on Synthetic, Test on Real; and Train on Real, Test on Synthetic) reveal a performance gap of only 3-10%, indicating strong distributional fidelity. Comparative analysis against classical oversampling methods -- SMOTE, Borderline-SMOTE, KMeansSMOTE, and SVM-SMOTE -- shows that QCBM achieves competitive classification performance and produces lower Maximum Mean Discrepancy (MMD) on the Telco dataset, suggesting superior structural similarity in certain imbalanced settings. These findings establish QCBM as a viable complementary tool for data augmentation, particularly for low-dimensional structured tabular data with class imbalance.

Image tracking is the problem of estimating the transformation that relates a moving image of a scene to an original reference image. The problem is important in control of autonomous vehicles or robots, where the image encodes information about the motion of the camera or environment, as well as in pure computer vision applications. In this paper, we present an equivariant filter design for high performance tracking of planar image transformations using an event camera. The design exploits the Asynchronous Event Blob (AEB) tracker (Wang et al., 2024) to extract feature-position measurements from the raw event stream, and an equivariant filter to compute an affine image translation and rotation using the special Euclidean group symmetry. The equivariant filter incorporates an equivalent-measurement update step that de-correlates the (highly temporally correlated) feature-position measurements provided by the AEB tracker. We evaluate the design experimentally using two datasets involving general and fast rotational motion. We benchmark results against direct optimisation (estimating the relative transformation from the raw blob tracks), and a covariance intersection approach for overcoming data correlation. Our design provides smooth image tracking for features moving up to 7000 pixels per second on the image plane.

The rapid growth of image data has produced large-scale datasets, raising concerns about the time and memory costs of model training. Selecting representative training subsets, however, remains challenging: individual sample contributions are unclear, and model behavior varies across datasets and runs. We address these challenges with a framework that combines coreset selection with an ensemble aggregation over multiple runs. For coreset selection, we propose SCOre-Stratified Selection (SCOSS), which partitions the training data into intervals based on a chosen score and samples from each interval. The ensemble combines predictions from multiple runs, each performed on an independently sampled training subset. As baselines, we use moderate and random selection, each in original and class-balanced versions. We assess the framework with Simple Graph Convolution (SGC) and Support Vector Machine (SVM) classifiers under different sampling ratios. Experiments show that SCOSS is competitive with baselines, often the best choice for SGC, and enables favorable trade-offs between accuracy and efficiency. On the fine-grained dataset, SGC with SCOSS outperforms SVMs when using fewer labeled samples. The code and supplementary materials are publicly available at http://scoss.lucasvalem.com.

While multi-agent debate (MAD) frameworks have shown significant potential in general reasoning, their effectiveness in highly structured, knowledge-heavy legal domains remains under-explored. In this work, we introduce the Legal Multi-Agent Debate (L-MAD) framework to systematically evaluate different debate structures and aggregation methods within Legal Textual Entailment. By assigning distinct expert personas to multiple agents, L-MAD improves upon strong single-agent baselines by up to 8\%. Furthermore, analyzing how debate scales reveals a clear trade-off: increasing the agent population reduces inconsistency and improves accuracy, whereas extending discussion rounds induces a detrimental \textit{over-deliberation drift} where agents reinforce each other's mistakes. Ultimately, our findings outline the practical boundaries and safety margins of deploying collaborative multi-agent systems in high-stakes legal reasoning environments.

Legal precedent retrieval is a fundamental task in legal case preparation, planning, litigation strategy, and legal research. Current approaches for automatic precedent retrieval map legal documents to a low-dimensional semantic space and compute similarity based on the proximity of their representations. These approaches treat legal documents as monolithic texts, ignoring the rhetorical organization of the legal technicalities. Ergo, they overlook nuanced legal meanings and fail to distinguish the contextual significance of legal entities and concepts that vary based on their rhetorical roles within the document. To address this insufficiency, we propose the PRecG pipeline that computes the similarity between pairs of legal judgments by hierarchically learning their representations. The process begins by decomposing each document into distinct semantic units (segments) based on the rhetorical roles of sentences. For each rhetorical segment, a knowledge graph is constructed to capture the legal entities and their relationships within the segment. Contextual representations of the entities are then learned and aggregated to derive segment-level embeddings. These embeddings are further integrated to produce a unified document-level representation, and finally, the semantic similarity between a pair of documents is computed. We validate the performance of the proposed approach through extensive experiments on a benchmark Indian legal dataset, comparing it against state-of-the-art baselines to demonstrate its effectiveness.

Knowledge graphs (KGs) are often automatically constructed from large-scale corpora, but they inevitably contain factual errors due to noisy sources and extraction failures, and verifying them reliably at industrial scale remains a critical challenge. To address this, we propose AgentKGV, the Agentic LLM-RAG framework for KG fact Verification, that integrates dynamic routing and iterative query rewriting, which handles surface-form mismatch in document-level retrieval. To make this framework more accurate and cost-efficient for industrial deployment, we further introduce a two-stage training strategy: turn-level distillation-based SFT that transfers reasoning ability from a large teacher model into a small model for stable query rewriting and reasoning, and trajectory-level GRPO that optimizes the search policy to reduce unnecessary retrieval at scale. On the long-tail-predicate split of the open-domain T-REx benchmark, our framework improves macro-F1 over single-turn RAG by 5.5 \%p, and two-stage training does it further by 9.4 \%p. GRPO also cuts the average number of search calls from 3.24 to 1.63 without lowering accuracy.

As audio-visual generative models evolve into world simulators, cross-modal synchronization stands as a critical proxy for assessing the consistency of world dynamics and causality in generated content. However, existing evaluation metrics presume structural correctness, reducing synchronization to mere temporal alignment. Consequently, they fail on generative outputs, especially when exhibiting structural hallucinations and asymmetric cross-modal relations, which currently \textbf{mandate expert human annotation to assess synchronization.} This dependency introduces a critical paradox: \emph{human evaluators rely on relative, reference-dependent comparisons, whereas automated metrics require reference-free, absolute scalars.} We resolve this paradox by proposing a framework that distills relative human perception into a continuous, globally consistent metric. First, we introduce SynthSync, a dataset of generative failures ranked via pairwise human annotations. Second, we adapt the Omni-LLM equipped with a continuous latent projection to translate relative human rankings into continuous absolute values. Third, we propose Real-Valued Group Relative Policy Optimization ($\mathbb{R}$-GRPO) to internalize the global causal structure of synchronization via listwise score distributions. Empirically, our metric achieves state-of-the-art human preference alignment. We leverage this estimator to establish a standardized benchmark, advancing AV-Gen assessment from low-level signal correlation to visually grounded causality.

The rapid expansion of large-scale AI models has led to significant performance breakthroughs across diverse domains, yet it has also raised critical concerns regarding computational costs, energy consumption, and environmental sustainability. This survey provides a comprehensive overview of the green development of large models, emphasizing resource-efficient architectures and full-stack hardware-software co-design. We systematically review recent advances in efficient model construction, including attention operator optimization, linear-complexity architectures, and model sparsification and merging, as well as training and deployment strategies such as data-efficient learning, parameter-efficient fine-tuning, and computational compression. Beyond algorithmic improvements, we explore energy-efficient AI hardware, including mainstream AI chips, memory optimization, cross-platform deployment, and sustainable infrastructure. Furthermore, we examine how large models are being applied to sustainability-critical domains such as DeepSeek, remote sensing interpretation, national-scale infrastructure, and global initiatives. Finally, we discuss key challenges and future directions, highlighting the need for continual learning paradigms, memory-centric hardware, and standardized evaluation protocols. This survey aims to offer a holistic roadmap toward sustainable, scalable, and socially responsible development of large models. Paper homepage: https://cje.ejournal.org.cn/article/doi/10.23919/cje.2025.00.438

Training-free in-context segmentation enables new object categories to be introduced at inference time from a single annotated reference image, eliminating the retraining and memory overhead of class-incremental learning. Recent approaches achieve this by combining vision foundation models for semantic correspondence with promptable segmentation networks like SAM. However, their performance is fundamentally limited by the quality of the cross-image similarity map; shared contextual backgrounds between the reference and query systematically elevate similarity in non-target regions, degrading prompt localization. We present REBASE, a training-free framework that explicitly suppresses these spurious contextual correspondences. Our method identifies the low-rank background feature subspace from the reference image and project the reference and query features onto its orthogonal complement in closed form, yielding cleaner semantic matching. We then generate positive point prompts using similarity-weighted farthest-point sampling, paired with a refined dense similarity prior. Without any training or parameter updates, our approach establishes a new state of the art among training-free methods on PACO-Part, FSS-1000, and cross-domain datasets such as ISIC2018, demonstrating that explicit background subspace removal is a highly effective principle for one-shot localization.

Although Video Large Language Models (Video LLMs) have shown strong performance in video understanding, their efficiency is still limited by the large number of visual tokens. Existing video token compression methods typically rely on frame-wise saliency or heuristic token merging, which can over-focus on locally salient regions and produce ambiguous fused features. To address these issues, we propose GeoTrace, a training-free spatiotemporal token compression framework that decomposes video evidence into exact skeleton tokens and traceable residual event tokens. Specifically, Contextual Farthest-Point Anchoring (CFPA) preserves salient, context-consistent, and high-coverage skeleton tokens, while Trajectory-Constrained Residual Condensation (TCRC) compresses residual tokens through one-to-one temporal trajectories and constrained near-manifold condensation, producing traceable event tokens with reduced ambiguity. We evaluate GeoTrace on four Video LLMs across four video understanding benchmarks, and the results demonstrate its effectiveness and generalization across different model architectures and scenarios. On LLaVA-OneVision, with only 10\% visual tokens retained, GeoTrace achieves a \(12.99\times\) TFLOPs reduction while preserving 99.1\% of the vanilla performance. Overall, GeoTrace offers a compact and traceable token representation for efficient and robust Video LLM inference. Code is available at \href{https://github.com/guohuan-xie/GeoTrace.git}{\texttt{Code}}.

Object detection is a fundamental component in numerous Unmanned Aerial Vehicle (UAV) applications, yet it has long been plagued by hindrances like occlusion or target pixel scarcity. Active Object Detection (AOD) provides a novel paradigm to address these challenges via active vision, while UAV-based AOD research remains scarce due to the lack of high-quality datasets and benchmarks for algorithm development and evaluation. To fill this gap, this paper presents ATRNet-LUDO, the first large-scale real-world dataset for UAV-Ground Active Object Detection (UGAOD). It contains 121,000 multi-view panoramic multi-target aerial images and 1.21 million local single-target slices, covering 10 vehicle targets across 40 scenarios. It enables the construction of diverse training and testing environments for UAV agent interaction and active observation policy learning. Based on this dataset, we establish a comprehensive evaluation benchmark for AOD policy learning methods. Most existing AOD policies rely on Deep Reinforcement Learning (DRL) but suffer from poor generalization. Evaluations on our benchmark reveal a significant generalization gap between training and testing performance, highlighting an urgent need for solutions. To this end, we leverage the Joint Embedding Predictive Architecture (JEPA) to construct a world model that enhances state representation learning, and propose AOD-JEPA by incorporating AOD-specific prior knowledge. Extensive experiments validate its effectiveness and superiority. We hope ATRNet-LUDO and the benchmark will advance research in the UGAOD field. The dataset and code are soon available at https://github.com/Leo000ooo/LUDO_dataset.

Cyberattacks on operational technology are increasingly causing costly downtime and physical damage, exposing the limitations of traditional rule-based monitoring in industrial IoT environments. While Large Language Models (LLMs) have strong semantic reasoning abilities to assist in decision support, their hallucinatory nature presents unacceptable safety liabilities for closed-loop control. This paper introduces a neuro-agentic control framework, a novel architecture that couples an LLM-based planner (i.e., such as Gemini 2.5 Flash-Lite) with a pre-trained Time-Series Foundation Model (TimesFM), to achieve physics-grounded autonomous defense. The paper introduces a ``Counterfactual Physics Injection'' mechanism that simulates the impact of LLM-proposed interventions within the numerical latent space of the foundation model before actuation, while allowing the system to reject hallucinatory or unsafe actions. Evaluated on an industrial dataset (e.g., the Secure Water Treatment (SWaT)) in the context of stochastic attack scenarios, the framework exhibited better performance compared to LSTM and TCN baselines. The Neuro-Agentic Loop prevented five breaches (33.3%) below the threshold versus LSTM (26.7%) and TCN (13.3%), with zero physically invalid (hallucinated) actions executed. These results demonstrate the efficacy of using foundation models as deterministic ``Sentinels'' to safeguard agentic AI in critical infrastructure.

Recent advancements in LVLMs necessitate robust benchmarks for complex, visually grounded reasoning. A critical limitation is identified in many document understanding benchmarks: visual content is often reducible to text, enabling high performance without genuine visual grounding. To address this limitation, OmniMapBench is introduced to foster visual-centric reasoning for map documents. The benchmark comprises 2,096 manually annotated question-answer pairs across 1,603 map documents from nine categories. It is designed to probe a hierarchy of skills, ranging from perception to multi-step visual reasoning. To quantify benchmark properties, a simple yet effective benchmark-level metric is proposed: the Visual Dependency Index (VDI), defined as the accuracy drop when images are replaced with question-agnostic descriptions. OmniMapBench exhibits higher VDI than established benchmarks, which quantitatively validates its focus on irreducible visual reasoning. Comprehensive evaluations of 25 leading LVLMs are conducted on OmniMapBench. A significant performance gap is observed, with the top-performing model achieving only 75.03\% accuracy. This result underscores the challenges posed by OmniMapBench to current LVLMs. This work aims to catalyze progress in visual-centric reasoning for document understanding of LVLMs. The dataset and code are publicly available at https://github.com/SIGMME/OmniMapBench.

Software engineering (abbrev. SE) has continuously evolved through increasingly powerful forms of reuse, from source code and libraries to components and services. Recent advances in AI agents have introduced a potentially new reusable artifact: skills. Emerging agent skill repositories and marketplaces enable developers to package, share, and reuse SE expertise as reusable skills. This trend raises a fundamental question: what SE activities are being encapsulated into reusable skills? Existing studies primarily focus on a broad range of skills acquisition, safety, or benchmarking, while lacking a systematic understanding of SE-specific skills and their coverage across the software development lifecycle. To address this gap, we conduct the first large-scale empirical study of SE skills in public repositories and marketplaces. We collect and analyze a large corpus of SE skills, examining the activities they encapsulate, lifecycle coverage, evolution characteristics, and evaluation mechanisms. Our findings reveal that SE activities are increasingly becoming reusable artifacts via skills and suggest promising research opportunities for skill recommendation and engineering-oriented structuring, as well as the need for mechanisms to encapsulate high-context SE activities into reusable skills. Overall, our study provides the first activity-centric characterization of SE skills and reveals how SE activities are increasingly being transformed into reusable skills. These findings offer new insights into skill reuse, ecosystem development, and the future of agent-centric SE.

Edge devices are increasingly utilized for deploying deep learning applications on embedded systems. The real-time nature of many applications and the limited resources of edge devices necessitate latency-targeted neural network compression. However, measuring latency on real devices is challenging and expensive. Therefore, this letter presents a novel and efficient framework, named EvoLP, to accurately predict the inference latency of models on edge devices. This predictor can evolve to achieve higher latency prediction precision during the network compression process. Experimental results demonstrate that EvoLP outperforms previous state-of-the-art approaches by being evaluated on three edge devices and four model variants. Moreover, when incorporated into a model compression framework, it effectively guides the compression process for higher model accuracy while satisfying strict latency constraints. We open source EvoLP at https://github.com/ntuliuteam/EvoLP.

Multi-UAV exploration is often constrained by unreliable communication, limited field-of-view sensing (e.g., lightweight onboard camera), and finite travel budgets that require each robot to reserve enough budget to return to its base. We present Dec-MARVEL, a decentralized budget-aware exploration framework for communication-free teams with directional sensing. Rather than exchanging maps, goals, or messages, each robot coordinates through its incidental observations: any teammate trajectory within its field of view serves as a coordination signal. A graph-attention actor fuses local frontier geometry, teammate motion, and budget features to select return-feasible waypoint-heading actions. The actor is trained with phase-conditioned critics, a training-only task-oriented privileged critic, and a mixture-based budget curriculum. Across 900 held-out trials spanning three team sizes (2, 4, 8 robots) and three travel budgets (720, 800, 1024 meters) against four baselines, Dec-MARVEL achieves the highest or tied-highest exploration rate and lowest sensing overlap across all nine team-size budget configurations. Under our tightest 720m budget, it reaches 53%, 94%, and 100% success for 2, 4, and 8 robots, versus 37%, 83%, and 99% for the strongest baseline. Physical-robot experiments demonstrate successful sim-to-real transfer and real-world deployment of Dec-MARVEL.

We present ARCANA, a collaborative multi agent framework for solving ARC AGI 2 tasks under strict test time and hardware constraints. ARCANA decomposes each task into iterative perception, hypothesis generation, symbolic execution, and reflective refinement. A perceptual grounding agent builds object centric scene graphs from raw grids, a latent program policy proposes diverse DSL programs, a symbolic executor verifies candidates on demonstrations, and a reflective agent synthesizes failure driven feedback for the next turn. These agents communicate through a shared differentiable blackboard and are scheduled by a learned meta controller. The design combines structured program search with adaptive multi turn correction, improving reasoning efficiency and solution quality on challenging abstract transformation tasks.

Cross-view geo-localization (CVGL) aims to achieve GPS-free localization by matching drone-view images with corresponding satellite-view images. Existing supervised methods rely on large-scale manually annotated cross-view image pairs, making them costly and difficult to scale. In contrast, existing unsupervised approaches typically depend on generative models or clustering-based stage-wise optimization, which are prone to distribution bias and the accumulation of noisy pseudo-labels. To address these limitations, we propose STEAM (Stable Self-Training with Elastic Matching and Adaptive Purification), an end-to-end unsupervised cross-view geo-localization framework that performs self-training directly on real drone and satellite images. Specifically, the proposed Stable Spatial-Aware Module enhances the stability of feature representations, Elastic Matching discovers high-quality cross-view pseudo-labels, and Adaptive Purification dynamically maintains a reliable pseudo-label repository throughout the self-training process. Extensive experiments on the University-1652 and SUES-200 benchmarks demonstrate that STEAM achieves state-of-the-art performance among all existing unsupervised methods and delivers performance comparable to supervised approaches, validating the effectiveness and superiority of the proposed framework. The source code is available at https://github.com/wsx-heu/STEAM.git.

Recent work has reported Emergent Misalignment (EM), where language models fine-tuned on narrow, domain-specific misaligned datasets abruptly acquire broadly misaligned behavior, alongside evidence that this behavior can be reversed through limited realignment. We systematically study repeated alignment and misalignment cycles using controlled fine-tuning loops while tracking behavioral performance, and LoRA representations throughout training. Although we reproduce EM, we find that both misalignment and realignment are highly sensitive to superficial dataset characteristics, with apparent rapid realignment largely disappearing after controlling for response-length differences. We further find that previously reported mechanistic signatures, including representational phase transitions in LoRA space, do not consistently correlate with behavioral misalignment across training. Our results suggest that current evidence for EM is less robust than previously claimed and highlight the need for evaluation protocols that carefully control for these surface level dataset artifacts to identify the robustness of the EM phenomenon.

Block sparse attention is a hardware friendly way to alleviate the key-value (KV) cache read bottleneck in large language models (LLMs). However, it is not prevalent among leading open-weight LLMs, which rely instead on dense attention or fine-grained selection, thereby motivating our analysis. We study DeepSeek's Native Sparse Attention (NSA) as a representative method, whose three-branch design lets us isolate block selection, the most challenging and consequential stage. We formalize selection and reduce it to ranking blocks by a single quantity, the attention mass: the sum of a block's attention scores. We show that if selection retrieves the blocks with the largest attention mass, block sparse attention can match the quality of dense attention. However, computing the exact attention mass requires reading every key, so the problem of block selection ultimately reduces to approximating this mass from a compact summary instead of the full keys. Via a cumulant expansion, we show why existing methods falter: their selection strategies attempt to estimate the attention mass, but are confined to a first-order approximation. Therefore, we propose COBS (Cumulant Order Block Sparse Attention), an attention method that builds on NSA, incorporating a novel selector that stores a compressed second-order statistic per block. On the 32k RULER long-context retrieval benchmark, COBS raises the NSA baseline's mean score from 0.2999 to 0.8195, approaching dense attention at 0.9040 and closing about 86% of the gap, while using only 1.21x the KV cache read traffic of the NSA baseline and 15.15x less read traffic than dense. The same model preserves short-context behavior and attains lower position-wise negative log-likelihood (NLL) than dense attention in our comparison.

Frontier autonomous-driving models -- especially vision-language-action (VLA) models, whose forward pass approaches $\sim$60~TFLOPs -- are outgrowing economical onboard deployment, since peak hardware sits idle most of the day. Cloud inference can instead share GPUs across active vehicles, but the vehicle must upload through a capacity-limited uplink, reach a GPU without queueing, and return a decision within the closed-loop budget. This paper asks: can the cloud drive? We answer with an analytical framework coupling communication limits, a roofline GPU service model, stochastic latency, and utilization-aware cost across three model classes, three offloading strategies, and three communication generations, applied to New York City. Separating a reactive 100~ms budget from a 300~ms deliberative tier (presuming an onboard reactive fallback), we find three \emph{nested} binding regimes. Communication binds first in dense cells: 5G fails early, 5G-Advanced is the practical threshold for feature-level offloading, and 6G adds headroom. Compute binds next under the reactive budget: near-term VLA is latency-infeasible regardless of bandwidth, because autoregressive FP16 decode is memory-bandwidth-bound (~114 ms on 2025 hardware). Its floor clears 100 ms around 2027; 6G then admits feature-level VLA by ~2028, 5G-Advanced only at light loading and not the dense corridor, and the deliberative tier from 2026. Cost binds last: once admissible, utilization-pooled cloud GPUs undercut onboard hardware for VLA, whose baseline (up to \$8,500 per vehicle-year) is expensive and idle; feature-level offloading (S2) is where the VLA cost crossover concentrates. Latency decides which model is admissible in which year; cost decides whether it is economical.

Reinforcement learning (RL) is increasingly used to post-train vision-language-action (VLA) models, but every update consumes robot rollouts that are slow and costly to collect, making sample efficiency a central concern. Manipulation tasks typically provide only sparse rewards, so a weak policy fails almost every rollout early in training and has little to learn from, even when those failures execute coherent behavior. Such a failure, however, is a success at a different task. We present Learning from Hindsight (LfH), which brings hindsight relabeling to RL post-training of VLAs by scoring failed rollouts against the tasks they actually achieved. A single vision-language model relabels both the instruction and the reward, proposing a hindsight instruction for a group of failed rollouts and scoring how well each satisfies it, and the policy trains on the relabeled and original rollouts jointly. Because VLAs generalize across language, relabeling in language lets the policy learn more from the same trajectories. On out-of-distribution LIBERO-PRO tasks, where standard RL improves only slowly, LfH achieves $5\times$ improvement in sample efficiency, and outperforms a dense progress-reward baseline. The gains hold across VLA backbones and on a physical Franka robot.

Vision-Language Models (VLMs) have achieved success using homogeneous Transformers to process multimedia data. Recent studies show that heterogeneous structures interleaving efficient mechanisms, like linear attention, improve both performance and inference latency over homogeneous designs. However, these efforts rely on handcrafted static mixing patterns, which are sub-optimal and difficult to adapt to specific hardware. To bridge this gap, we propose Multi-Objective Search for Adaptive Inter-layer Composition (MOSAIC), a hardware-aware search method that automatically transforms homogeneous models into optimized heterogeneous architectures. MOSAIC integrates diverse efficiency mechanisms--including linear, sparse, and low-rank operators--into a unified search space. By formulating the selection as a multi-objective Mixed Integer Programming (MIP) problem, our method identifies optimal configurations that maximize downstream performance under strict hardware latency constraints. To mitigate performance degradation from structural transitions, we introduce a two-stage parameter recovery process: global off-policy distillation to stabilize internal representations, followed by a dual-teacher on-policy distillation leveraging a 235B oracle for knowledge expansion and the original 4B teacher for distributional stability. We validate MOSAIC through MOSAIC-4B, derived from Qwen3-VL-4B-Instruct. Results demonstrate that MOSAIC-4B matches the baseline's performance across multiple benchmarks while requiring less than 2% of the original training cost. Furthermore, it substantially improves inference efficiency, achieving 1.76x prefilling and 2.54x decoding speedups.

Driven by next-token prediction, NLP shifted from task-specific models into powerful generalist foundation models. What, then, is the equivalent catalyst needed to achieve a general-purpose model in computer vision? In this paper, we contend that large-scale text-to-video generation serves as a strong pre-training paradigm for computer vision, providing the necessary spatiotemporal priors, vision-language alignment, and scalability required for general visual intelligence. We introduce GenCeption, which leverages a pre-trained video generative diffusion backbone to define a feed-forward perception model, capable of performing various vision tasks steered by text instructions. Empirical results demonstrate that GenCeption achieves state-of-the-art performance across a diverse suite of tasks, including depth, surface normal, and camera pose estimation, expression-referring segmentation, and 3D keypoint prediction, often matching or surpassing specialized models (e.g. DepthAnything3, SAM3, D4RT, VGGT-Omega, Sapiens, David, Genmo, and Lotus-2). Furthermore, the video generative pretrained backbone outperforms alternative pretraining paradigms (e.g., V-JEPA, and Video MAE) under comparable settings. Importantly, GenCeption exhibits preliminary data and model scaling properties along with exceptional data efficiency, where it achieves comparable performance with leading models like D4RT and VGGT-Omega with 7 to 500 less training data. Finally, GenCeption also exhibits intriguing emergent behaviors: a model trained exclusively on synthetic human videos generalizes to real-world footage and out-of-distribution object categories (e.g., animals and robots). These findings suggest that video generation is not merely a synthesis tool, but a foundational path toward generalist vision intelligence for the physical world. Project page: https://genception.github.io

We study contextual bandit problems with correlated arms and access to surrogate reward signals produced by a machine learning model, motivated by applications such as large language model (LLM) routing. Unlike classical contextual bandits that rely solely on bandit feedback and assume conditional independence across arms, our setting allows context-dependent inter-arm correlations and auxiliary reward information that may be noisy or misspecified. We propose algorithms that leverage such surrogate rewards through two complementary designs. A coupled reward-mixing approach pools true and surrogate rewards to accelerate learning when surrogate signals are reliable, while a decoupled prediction-mixing approach maintains separate estimators for bandit feedback and surrogate rewards and adaptively combines their predictions. This decoupling yields robustness to surrogate misspecification, recovering regret guarantees comparable to reward-only bandit methods in the worst case, while achieving improved regret when surrogate predictions are sufficiently informative. We provide theoretical regret analyses for both approaches and evaluate them on LLM routing benchmarks under varying accuracy versus cost trade-offs. The results demonstrate improved sample efficiency and consistently better accuracy-cost trade-offs compared to standard contextual bandit baselines and strong static routing methods.

Exposed documents such as emails, chat threads, tickets, and incident notes routinely leak credentials, but during incident response a leaked secret is only half the story. Responders also need to identify the ``door'' the secret opens: the account, tenant, endpoint, database, cloud resource, or other system that the credential could allow an attacker to access. Traditional secret scanners rely on regular expressions or trained classifiers which work well on well-formatted code, yet they struggle when a credential is fragmented, reformatted, or far from the resource it unlocks, and they report the secret string without naming what it opens. We present Secret Scanner Agent (SSA), a multi-agent large-language-model system that extracts both the secret and its associated door, together with supporting evidence, from unstructured exposed documents. SSA pairs a detection agent that favors recall with a review agent that filters false positives and recovers missing context. Because real credential data is sensitive, we evaluate SSA on synthetic benchmarks we generated that span 23 secret types and multiple document formats, scored with a three-step pipeline of programmatic matching, an LLM judge, and human review. Across six models, multi-agent SSA improves extraction precision over a single-agent variant, with the largest gains on door extraction, by up to 16 percentage points. SSA matches a regular-expression scanner's precision while more than tripling its recall, and against thirteen security analysts it is more precise, recovers nearly twice as many secret--door pairs, and runs five to seventeen times faster. By returning the secret, its door, and supporting evidence in one result, SSA turns credential detection into an actionable finding for triage and remediation.

Safety-critical perception systems must reliably detect rare object classes within small label spaces, a setting that long-tailed detection methods, designed for hundreds of classes with dense annotation, fundamentally do not address. Open-vocabulary detectors offer a promising alternative, as they use natural language queries at inference time, making prompt quality a first-class lever for detection performance. We exploit this property to address class imbalance: rather than retraining models or collecting additional annotations, we ask whether iteratively refining the language prompts, fed to frozen detectors, can improve minority class detection. We introduce C-GAP Caption-Guided Augmentation and Prompting), a detector-agnostic, annotation-free framework that operates in two phases. First, we establish a composite caption baseline combining per-image scene descriptions with class-quantity context, which we show outperforms scene-description only or class-quantity-only prompts across multiple open-vocabulary architectures and benchmarks. Second, an LLM iteratively refines each image's caption individually, with trials triaged into accept, tentative, or regenerate buckets based on minority-class [email protected] against a dynamic threshold derived from the composite baseline. Refinement terminates early once sufficient [email protected] gain is achieved. No detector weights are updated at any stage. Our experiments shows that C-GAP improves minority-class average precision up to 53% over the baselines. On COCO, C-GAP improves minority-class [email protected] by ~81% relative over the composite baseline (17.69 -> 32.09). Experiments confirm that composite captions provide the critical foundation for effective refinement: using scene-description-only or class-quantity-only prompts as the refinement starting point yields diminishing returns, supporting both stages of C-GAP as necessary contributions.

Graph Neural Networks have emerged as a powerful tool for the fast and accurate prediction of various crystal properties. These models often encode domain-specific knowledge into their graph encoding modules, which increases their parameter size and makes their performance heavily dependent on domain expertise. Added to this, explicitly incorporating all chemical and structural features, that might influence a specific crystal property into the GNN encoder, is a challenging task. In this work, we propose a soft prompt learning framework that captures latent features essential for property prediction, which are not explicitly provided to the GNN. We introduce a novel multilevel graph prompt learning framework comprising both node-level and graph-level soft prompts. At the node level, we capture the local chemical semantics of different atom types, while at the graph level, we encode the global structural symmetry of the crystal graph. Our proposed prompt learning framework is lightweight and seamlessly integrates with any existing GNN encoder. Extensive experiments on popular benchmark datasets show that incorporating prompt learning significantly improves (3\% - 15\%) the performance of state-of-the-art GNN models in crystal property prediction tasks. Furthermore, the learned soft prompts enable cross-property knowledge transfer, enhancing prediction performance for properties with limited training data. Code is available at https://github.com/shrimonmuke0202/Prompt.git

We formalize a research result in the Lean 4 proof assistant by having a mathematician direct an AI system, and frame the activity as a formalization game. The objective is to turn a LaTeX document into Lean. The game is won when the development compiles, contains no sorry, and a machine check shows the target theorems rest on Lean's foundational axioms alone. Reuse is a second check, by a definition we introduce: whether the development yields a self-contained layer of general mathematics the wider library could absorb. The case study is a complete, axiom-clean formalization of well-posedness for the nonlinear Vlasov equation via Dobrushin's mean-field route -- existence, uniqueness, the stability estimate and mean-field limit, and a short-window superposition principle (weak solutions are Lagrangian). The human's role was to direct, not to write proofs: to scope the definitions, steer the decompositions, and triage the library's gaps; the AI agent executed. The formalization certifies the proof of each statement as written; whether the written statement is the intended theorem stays the mathematician's judgment. The optimal-transport machinery that fell out of the build (in particular, properties of the Wasserstein-1 metric and the Kantorovich-Rubinstein duality theorem) separates into a self-contained layer that compiles against Mathlib alone: about a sixth of the development (49 of 299 declarations), behind a 22-declaration interface with no reverse dependency. The headline theorems ran in about a week, the full development in about a month. We report the quantitative claims as observations of one game, not as general laws. The game's rules name no particular system, so the methodological framing is meant to outlast the tools of any one run.

AlphaZero has demonstrated that a neural-guided Monte Carlo Tree Search can achieve superhuman performance, but strong play does not necessarily imply perfect play. We study this gap in two oracle-evaluable domains with contrasting structure: Connect Four, a solved partisan game with exact game-theoretic values, and Chomp, an impartial game whose optimal play is governed by Grundy-number structure. Under a unified self-play $+$ MCTS pipeline, we compare vanilla AlphaZero, a multi-frame variant (limited to Chomp), and an AlphaZero Auxiliary Loss (AZAL) that adds oracle-derived policy supervision. We find that vanilla AlphaZero achieves strong play across both domains but cannot preserve the exact trajectories required for optimal play: in Connect Four, it fails to maintain the optimal line of play, while in Chomp, it fails to consistently restore the $g=0$ invariant. On rectangular Chomp boards, multi-frame inputs alone do not remove this gap. Nevertheless, AZAL substantially improves oracle consistency across multi-seeded full-game traces and sampled-state evaluations. On Chomp, AZAL reaches perfect full-game oracle consistency on 10x11 and high but not complete consistency on 9x10; on Connect Four, AZAL improves oracle-match rate and delays the first oracle mistake, but does not reach perfect play.

While autonomous coding agents have significantly advanced automated test generation, they remain fundamentally limited by lazy generation, a phenomenon where agents prematurely terminate tasks and systematically avoid complex programmatic logic, resulting in inadequate code coverage. Currently, mitigating this premature termination requires continuous human-in-the-loop supervision. This heavy reliance on human intuition creates a bottleneck that negates the efficiency gains of automated generation. We propose SCATE, a framework for adaptive, automated supervision of coding agents that replaces human intervention during test generation. By formulating supervision as a contextual bandit problem, SCATE learns to select the most promising testing actions based on the current coverage and class testability metrics, maximizing coverage gains while minimizing wasted generation effort. Our empirical evaluation demonstrates that SCATE integrates seamlessly with different coding agents. When applied to GEMINI-CLI, it achieves 32.3% higher line coverage and 30.9% higher branch coverage than the agent-only baseline. A comparison with CLAUDE CODE confirms the framework dynamically adapts its policy to optimize each agent's unique strengths. SCATE also consistently outperforms state-of-the-art non-agentic approaches across all metrics.

Distributed IoT systems generate multivariate time-series streams for monitoring physical assets, servers, and embedded sensing platforms. Detecting abnormal temporal behavior is critical for fault diagnosis, predictive maintenance, and security. However, practical IoT anomaly detection is hindered by decentralized and non-IID data, limited bandwidth, and the constrained computation and memory of edge devices. This paper proposes FedKAD, a resource-efficient federated Koopman anomaly detection framework for distributed IoT multivariate time series. Unlike deep-learning-based anomaly detectors that require training and communicating large neural models, FedKAD learns normal temporal dynamics through lightweight sliding-window Koopman representations. Federated training is formulated as a low-rank consensus problem, where raw sensor streams and local reduced dynamics remain on device while only compact subspace variables are exchanged with the server. To optimize the shared representation under orthonormality constraints, we develop a federated Stiefel-ADMM algorithm and provide convergence and stationarity analysis under partial client participation. During inference, each client detects anomalies locally by measuring the prediction residual between observed future trajectories and the learned Koopman dynamics. Experiments on four widely used multivariate time-series anomaly detection benchmarks show that FedKAD maintains or improves detection performance compared with federated deep-learning baselines. More importantly for IoT deployment, FedKAD provides up to $2.1\times10^3$ faster training, $80\times$ lower communication, and $79\times$ lower inference latency than neural baselines, confirming its suitability for resource-constrained edge devices.

Vision-language-action models (VLAs) inherit semantic capabilities from pretrained VLMs, yet large-scale post-training on robot data and architectural modifications can reshape the backbone so extensively that it becomes difficult to isolate what the VLM contributes to control. Directly converting pretrained VLMs into VLAs with minimal architectural change offers a more transparent path to understanding how VLM capabilities transfer across model scales. The core obstacle is output-distribution mismatch: predicting actions as bare numeric token sequences moves generation away from the VLM's pretrained language distribution, degrading the capabilities we seek to preserve. To address this, we propose CLAP (Causal Language-Action Prediction), which prepends each numeric action sequence with a natural-language action description, causally conditioning precise action-token prediction on a language-action plan without modifying the backbone architecture. With single-epoch fine-tuning alone, 2B CLAP achieves 90.8% on LIBERO (+14.9 pt over VLA-0) and improves robustness on LIBERO-PRO under language, object, and spatial perturbations. We will release CLAP at 0.8B, 2B, and 4B as an open-weight, multi-scale compact VLA family from a single VLM lineage, enabling controlled analysis of VLM-to-VLA capability transfer.

The stochastic linear bandit, where actions are represented as vectors and rewards are linear, is a central paradigm for sequential decision making. We study a partially observed variant of this problem in which the learning agent only sees a random subset of coordinates for each action. Such partial observability arises naturally in settings like recommendation and healthcare, where full action descriptions can be expensive or even impossible to obtain. In general, this makes sublinear regret information-theoretically impossible. However, we show that this barrier can be overcome when the action vectors have low intrinsic dimension. We propose an algorithm, TOFU-POV, that estimates the latent action subspace using the masked actions, imputes current actions using an epoch-wise frozen representation, and runs OFUL in the resulting low-dimensional coordinates. Our theory shows that TOFU-POV enjoys a $\sqrt{T}$ regret that scales with the intrinsic action subspace dimension as opposed to the ambient dimension and quantifies the interaction between these quantities and the missingness, decision set size, and subspace conditioning. We also devise a rank-adaptive algorithm that does not require the knowledge of the intrinsic dimension. We complement these guarantees with a lower bound based on a novel product construction that separates usual reward-learning uncertainty from a missingness-dependent cost intrinsic to partial observation. Synthetic and real data experiments support our theory and show that TOFU-POV can substantially improve upon natural baselines in this challenging problem.

Recent benchmarks for VLMs largely assess single- or limited-view perception, leaving untested the core cognitive ability to integrate observations across viewpoints into a coherent, world-centric (allocentric) 3D mental model. We introduce MultiView-Bench, a diagnostic benchmark expressly designed to evaluate multi-view integration for holistic 3D scene comprehension. Unlike existing datasets that focus on pixel-level mapping or camera-relative navigation, MultiView-Bench requires models to decouple object positioning from transient perspectives and ground them in a fixed global coordinate system. This capability serves as a prerequisite for VLMs before being deployed for downstream tasks such as mechanical part assembly. Our systematic evaluation of frontier VLMs reveals consistent failure modes: strong performance on 2D planar relations from a single image, but marked difficulty with 3D spatial relations and with aggregating information across views. We further identify biases in VLMs, such as struggles with unconventional axis directions and sensitivity to object colorways and texture variations. Acknowledging these limitations, we propose ViewNavigator, a multi-agent framework that actively selects informative viewpoints, perceives, and fuses multi-view evidence, improving diverse base models on MultiView-Bench even under a strict budget-matched comparison (and by 3-5x for the full agent).

Soft metamaterials provide a promising platform for robotics, biomedical devices, and flexible electronics. The localized mechanical responses by nonuniform excitation are ubiquitous in soft materials, yet their controlled transmission across assemblies remains largely overlooked in metamaterial design, which critically constrains nontrivial functionalities with end-to-end and long-range deformation transmission. Here, we introduce an impedance-guided design framework that enables programmable transmission of localized deformation in modular soft metamaterials, achieving behaviors unattainable by intuitive design. By establishing a nonlinear model considering position-dependent interactions and integrating the concept of mechanical impedance within metamaterials, we regulate assembly-level transmission solely through unit-cell topology optimization. The resulting framework enables effective synthesis of module families, allowing both homogeneous and heterogeneous assemblies to be custom-built with markedly enhanced transmission characteristics. Leveraging the highly combinatorial and extensible design space, we physically realize diverse on-demand displacement manipulation architectures, including obstacle-bypassing modular soft-metamaterial assemblies, defect-tolerant soft gripping, and embodied signal processing. Beyond deformation programming, the reconfigurability and reassemblability of these soft modules can embed electric logic signals, enabling energy-efficient and low-latency information processing through compliant-switch-controlled mechanical LED displays and wearable finger-motion-sensing controllers. Our method provides fundamental insights into localized deformation transmission in modular soft metamaterials and establishes a scalable route toward embodied-intelligence material systems, particularly for soft-metamaterial-centric actuation, sensing, and collective computing.

AI agents have become capable of autonomously completing short, well-specified tasks. However, existing terminal benchmarks largely focus on simple problems that finish within minutes and are evaluated only by their final outcome. This setup overlooks intermediate progress and partial solutions, yielding sparse reward signals and an incomplete picture of agent capability. We introduce Long-Horizon-Terminal-Bench, a terminal benchmark of 46 long-horizon tasks spanning nine categories, including experiment reproduction, software engineering, multimodal analysis, interactive games, and scientific computing. Each task follows a Terminal-Bench-style setup with a reference solution or simulation engine, but is further decomposed into fine-grained graded subtasks. This design enables dense intermediate rewards and partial credit, allowing evaluation to capture not only whether an agent reaches the final goal, but also how far it progresses on open-ended workflows. Tasks in Long-Horizon-Terminal-Bench typically require hundreds of episodes and minutes to hours of execution, stressing long-horizon planning, long-context management, and iterative debugging rather than one-shot problem solving. We evaluate 15 frontier models and find that agents consume on average 9.9M tokens per task, with roughly 231 episodes and 85.3 minutes of execution time per run, making Long-Horizon-Terminal-Bench more demanding than prior terminal-based benchmarks. Even the strongest tested model achieves 15.2% pass@1 at a partial-reward threshold of 0.95 and 10.9% at a perfect-reward threshold of 1.0, while the mean pass rate across models is 4.3% and 1.7% under the two thresholds, respectively. These results reveal headroom for improvement. We further analyze failure modes and error patterns, and release Long-Horizon-Terminal-Bench to support future progress on long-horizon terminal agents.

Warehouse operations are governed by Standard Operating Procedures (SOPs) that encode complex, multi-system decision logic, which must be executed reliably under strict time constraints, yet LLM agents lack mechanisms to enforce procedural compliance and degrade under the context overload full SOP specifications introduce. We present Eluna, a production-deployed agentic system for reliable SOP execution. Eluna is a graph-guided, multi-agent framework that encodes SOPs as directed acyclic graphs with progressive disclosure and delegates independent tasks to parallel sub-agents, each with persistent code execution and live data access. To meet production latency and accuracy needs, we use asymmetric episodic distillation where a strong teacher is improved through episodic error memories, then a smaller student is fine-tuned on the corrected trajectories with memory stripped, internalizing corrections without inference-time overhead. On a 13-task benchmark and two production applications, our fine-tuned models match or exceed their teacher, beat all larger off-the-shelf baselines, and reach 94% expert agreement on the ticket processing application.

Robotic manipulators excel in structured environments but face substantial challenges in unstructured and dynamic settings. This paper presents SplatCtrl, a unified framework for real-time scene reconstruction and reactive robot motion generation to enable collision-free robotic arm control in previously unseen and continuously changing environments. Building on 3D Gaussian Splatting (3D-GS), we introduce a hybrid voxel-based filtering and dynamic Gaussian relocation strategy that supports efficient scene reconstruction from RGB-D streams while accommodating environmental changes. For safe and reactive control, we further propose a method for deriving continuous signed distance functions from isotropic Gaussians, providing stable and differentiable collision probability estimates that bridge classical distance fields with the modern implicit representation. These continuous distance metrics are incorporated into control barrier functions, resulting in a unified perception-action coupling framework that supports smooth and reliable real-time motion generation in response to scene changes. Experimental validation in simulation, on physical robot, and within shared human-robot workspace demonstrates the framework's effectiveness, achieving integrated scene reconstruction and reactive control in uncertain, and dynamic environments.

Efficient serving of diffusion large language models (dLLMs) is hindered by convergence heterogeneity: when batching multiple requests, different sequences converge at different rates, causing faster requests to stall behind slower stragglers and introducing compute bubbles and tail latency. We present BlockServe, a continuous batching framework that integrates block-grained scheduling -- immediately evicting completed requests at block boundaries -- with mixed-state execution that extends dual cache and parallel decoding to heterogeneous batches via gather-scatter indexing. Furthermore, a compute-aware admission controller expands effective batch capacity through token-budgeted refill. On Dream and LLaDA across five benchmarks, BlockServe achieves 1.9--10.6$\times$ throughput over Fast-dLLM with comparable generation quality, establishing block-grained scheduling as a foundation for high-throughput offline dLLM inference.

Training reinforcement-learning agents directly on physical robots makes every fall costly, since a fall can damage the platform and cannot be undone like a simulator reset; the goal is therefore to minimize falls during training rather than trade them off against return, as constrained Markov decision process (MDP) formulations do. A standard mitigation hands control to a separate recovery policy whenever the agent leaves a designer-specified safe region (a subset of state space it should stay within), but the resulting mixed-policy rollouts silently bias every on-policy update, and the importance-sampling correction that would remove this bias is ill-defined whenever the recovery policy is deterministic. We address this bias with a drop-in modification of proximal policy optimization (PPO). Its core is an unbiased policy-gradient estimator that uses the score function only at safe timesteps and never evaluates the recovery policy's density, so it stays valid even when the recovery policy is deterministic, exactly where importance sampling breaks, and it empirically dominates importance sampling even when the recovery policy is stochastic. Because the recovery policy still makes credit assignment slow near the safe-region boundary, two further components accelerate learning: a closed-form value for recovery-triggering states when dynamics and recovery are deterministic, and an imitation loss that copies recovery actions only when recovery succeeds. On a three-environment, five-seed benchmark, the resulting algorithm reduces training-time falls by factors of 233x, 48x, and 26x on HalfCheetah, Ant, and Unitree Go1 over standard PPO, while matching or exceeding PPO's final reward, and on Ant, where the recovery policy is unreliable, it is the only method that reaches 80% of the best final reward.

We describe our entry to the ICIP 2026 Grand Challenge on Extreme In-the-Wild License Plate Super-Resolution (XLPSR), which scored 9.73 wECR on the public validation leaderboard. The system pairs a Hybrid Attention Transformer super-resolution (HAT) front-end with an ensemble of two scene-text recognisers (PARSeq-S and CLIP4STR-B) and a confidence-weighted character-voting scheme that abstains on uncertain positions. We treat XLPSR as a recognition task gated by image legibility: the SR step exists to lift characters out of sub-pixel territory, and the asymmetric scoring rule (+2 / -1 / 0) is exploited explicitly through abstention. Our pipeline runs in 1.7 s per sequence on RTX 3090 (max 2.7 s, p99 2.4 s), well under the 60 s/sequence Docker budget.

Large Language Model (LLM) agents have shown promise in multi-step planning tasks, but existing approaches like LATS (Language Agent Tree Search) and ReAct rely heavily on LLM inference during planning, leading to high computational costs and stochastic behavior. We present \textbf{GATS} (Graph-Augmented Tree Search), a planning framework that combines systematic UCB1-based tree search with a layered world model to eliminate LLM calls during inference while achieving superior planning performance. Our three-layer world model integrates: (L1) exact symbolic action matching, (L2) statistics learned from execution logs, and (L3) LLM-based prediction for unknown actions. On synthetic planning tasks with branching paths and dead-ends, GATS achieves \textbf{100\% success rate} compared to 92 % for LATS and 64\% for ReAct. On a comprehensive stress test spanning 12 challenging scenarios -- including coding workflows, web navigation, and long-horizon tasks -- GATS maintains \textbf{100\% success} while LATS drops to 88.9 % and ReAct to 23.9%. GATS requires \textbf{zero LLM calls per task} during planning (vs. 37 per task for LATS) and produces deterministic plans with zero variance across runs. Our results demonstrate that systematic search with learned world models can substantially outperform LLM-guided exploration for agent planning.

Computing Nash equilibrium policies in multi-agent Pursuit-Evasion games (PEG) is challenging due to the exponential growth of the joint state and action spaces with the number of agents. Existing approaches either rely on offline equilibrium approximations, which may lack adaptability during execution, or online planning methods, which suffer from large branching factors. In this work, we propose Primitive-Guided Tree Search (PGTS), a hybrid framework that integrates offline exact Nash equilibrium computation with online tree search: PGTS first solves a collection of smaller, tractable sub-games offline; at deployment, PGTS performs online tree search at each time step, using the optimal sub-game policies and value functions to guide tree expansion and estimate leaf-node values. Extensive experiments on varied graph topologies, including real-world networks, demonstrate that PGTS significantly outperforms state-of-the-art learning and heuristic baselines, while maintaining robust performance against adversaries.

Indoor scene relighting demands photorealism, precise spatial control, and strict multi-view consistency. While diffusion-based image editing models enable semantic lighting manipulation via text prompts, enforcing exact 3D light placement often disrupts their generative priors. We propose Lume-Palette, a progressive framework that leverages semantic lighting priors for spatially controllable multi-view indoor relighting. The approach decouples relighting into two stages: (1) illumination distillation, which extracts canonical illumination palettes from a pretrained diffusion model to preserve realistic material-light interactions, and (2) illumination casting, which explicitly maps target spatial lighting conditions defined from coarse 3D geometry. To efficiently handle dense multi-view and multi-modal inputs, we introduce an asymmetric multi-view conditioning strategy that selectively injects essential spatial context. Experiments on diverse synthetic scenes and real-world scenes demonstrate that Lume-Palette produces photorealistic, spatially controllable, and multi-view consistent relighting results. Project Page: https://cjeen.github.io/lumepalette

Pretrained generative robot policies based on flow matching and diffusion have achieved impressive results across a wide range of manipulation tasks. Yet real-world deployments routinely expose failure modes outside the pretraining distribution. Closing these gaps typically requires large-scale data collection or online reinforcement learning on physical hardware, which is impractical for rapid and safe adaptation. We present FlowDAgger, a sample- and compute-efficient method for adapting frozen generative robot policies from human interventions in latent space. Our key idea is action inversion: each human expert action is mapped to the noise that would have produced it under the frozen base policy, using reverse-time integration followed by local refinement. The resulting inverted noise provides supervision for a lightweight latent policy that steers the base model at deployment time, enabling rapid skill acquisition while preserving its behavioral priors. We evaluate FlowDAgger in simulation and on real-world bimanual and single-arm manipulation, adapting both action-head VLAs and world-action models from a handful of interventions. FlowDAgger outperforms supervised fine-tuning and latent-space RL baselines and preserves pretrained skills on held-out tasks, offering a practical path for adapting robot foundation models in the real world. Website: https://microsoft.github.io/FlowDAgger

Human egocentric video is a scalable supervision source for humanoid policy learning, but current pipelines struggle with hand-object occlusion, oversimplified motion, or specialized capture hardware. We introduce AgenticFocus, a Mixed Reality synthesis pipeline that converts ordinary first-person-view human videos into robot-trainable demonstrations by restoring occluded object geometry, reconstructing full-hand motion, and retargeting it to a humanoid embodiment through camera-relative alignment and layered compositing. The resulting dataset pairs focused visual observations with synchronized robot actions and states. AgenticFocus achieves lower trajectory error and smoother wrist motion than cross-embodiment baselines, with SPARC scores of -5.18 versus -5.56 and -6.05.

Multimodal Large Language Models (MLLMs) are typically designed under the assumption that all modalities available during training will also be accessible at inference. However, many real-world settings violate this assumption, requiring models to operate under a privileged modality setting, where auxiliary modalities are available only during training. While these modalities contain valuable information, existing MLLMs largely fail to leverage them effectively, as they treat modalities as interchangeable inputs rather than sources of complementary supervision. We propose Mixture of Probes (MoP), a novel framework that disentangles modality-specific and modality-general signals within the MLLM, allowing the model to preserve modality-dependent structure while learning transferable representations across modalities. At its core, MoP achieves this through a structured probing mechanism that extracts and organizes information from intermediate representations of a shared modality encoder, rather than relying only on final-layer alignment as done in existing MLLMs. To support this disentanglement, we further introduce MoP Cross-modal Training (MoP-X), a training strategy for MoP centered around a probe disentanglement loss that prevents probe collapse and encourages cross-modal learning. We evaluate MoP across two domains spanning eight tasks and four modalities under a comprehensive evaluation protocol tailored to the privileged modality setting, where each modality is independently treated as the sole input at inference time. MoP consistently outperforms strong MLLM baselines, achieving up to 65% relative improvement, demonstrating that auxiliary modalities, even when unavailable at inference, can provide substantial gains when effectively leveraged during training. Code, model checkpoints, and evaluation protocols will be made available at https://github.com/Sony/MoP.

Exploration is essential to RL since a policy cannot improve by repeatedly sampling the behaviors it already prefers. Standard methods inject stochasticity in the action space, but such jitter only yields rollouts close to the original. Escaping a weak policy often requires global perturbations that action noise cannot produce. Large language models (LLMs) and vision-language-action (VLA) models offer a pathway: they condition the policy on a natural language prompt, and since the rollout follows from it, modifying the prompt induces global changes. The challenge is finding prompts that induce useful global changes. With a weak policy that rarely succeeds, reward is too sparse to select on. Our idea is to refine prompts from the rollouts themselves: a vision-language model (VLM) reasons over the rollout video, diagnoses how the policy responded, and rewrites the prompt to elicit better behavior next time. This procedure realizes posterior sampling, a classical RL exploration framework, at the level of prompts: the VLM maintains an implicit distribution over useful prompts and updates it from observed rollouts. We call this strategy Prompt-Driven Exploration (PDE). Across manipulation and reasoning tasks, PDE enables RL to learn successful policies even from zero-reward starts, and improves sample efficiency more broadly. Our website is available at https://xinyunsunshine.github.io/prompt-rl.

Estimating 3D geometry in underwater environments presents unique challenges due to light attenuation, scattering, and the absence of large-scale, high-quality 3D annotations. Pioneering methods rely on massive dense annotations that are impractical in underwater settings. In this paper, we propose Wat3R, a cross-domain semi-supervised learning framework designed to adapt feed-forward 3D reconstruction models from air to underwater scenes. Uniquely, our method eliminates the need for any annotated underwater data following a teacher-student architecture, that learns robust geometry representations merely on abundant unlabeled real underwater video footage. We also design a cross-view consistency loss that leverages geometric cues from other views to compensate for the information degradation in the current view caused by water attenuation and scattering. Furthermore, considering the lack of comprehensive evaluation benchmarks, we construct Water3D, a diverse dataset covering various water bodies and underwater scenarios, designed for geometric task evaluation. Experimental results demonstrate that Wat3R outperforms current state-of-the-art methods in underwater multi-view depth estimation and point cloud reconstruction. The dataset and code are available at https://github.com/LSXI7/Wat3R .

Monocular depth estimation has seen remarkable progress through foundation models achieving robust zero-shot generalization, yet their computational demands place them far beyond the reach of embedded and mobile platforms. Lightweight alternatives exist, but have been developed almost exclusively within single-domain, self-supervised paradigms, failing silently under domain shift. We present ZipDepth, a compact monocular depth network that bridges this gap by combining an efficient reparameterizable encoder-decoder with large-scale knowledge distillation from a foundation model over a large multi-domain training set. Comprising just 6.1M parameters, ZipDepth runs at real-time rates from server GPUs to power-constrained devices, achieving the best trade-off between zero-shot accuracy and deployment efficiency among lightweight models across five benchmarks, taking a significant step towards the accuracy of foundation models with 50x more parameters.

Recovering high-quality video from sparse event streams is a challenging task. Regression methods often blur textures, while existing generative models struggle with long-term stability. We propose LongE2V, a novel approach that leverages pre-trained video diffusion priors to jointly handle event-based video reconstruction, prediction, and frame interpolation. By fine-tuning a foundational video model, our approach achieves high data efficiency and superior perceptual quality. We introduce Autoregressive Unrolling and Adaptive Context Switching to mitigate temporal drift in extremely long sequences. We also propose Reencoding Alignment with Cross Residual Correction to ensure precise bidirectional consistency during frame interpolation. Furthermore, Event Voxel Density Augmentation ensures robustness across varying sensor resolutions. Extensive experiments on real-world benchmarks demonstrate that LongE2V outperforms state-of-the-art methods across all three tasks, exhibiting exceptional temporal coherence and zero-shot generalization. Project page: https://cdfan0627.github.io/LongE2V-page/

Scaling 3D Gaussian Splatting (3DGS) to large outdoor scenes is costly in both data acquisition and computation. Adopting panoramic images with equirectangular projection (ERP) can reduce capture effort via their full $360^{\circ}$ field of view, yet the resulting omnipresent visibility invalidates existing partitioning strategies that rely on local camera frustums, causing block-wise optimization to degenerate into global training. Thus, we propose PanoLOG, a two-stage coarse-to-fine framework equipped with a Geometry and Gradient-based Partitioning Strategy tailored for large-scale panoramic 3DGS reconstruction. In the global coarse stage, PanoLOG leverages sky-sphere modeling and panoramic monocular depth supervision for reliable geometry, while in the refinement stage, G$^2$PS builds adaptive bounding volumes via parallax-driven uncertainty and assigns cameras via gradient-based importance scoring. Furthermore, we construct Pano360, the first benchmark on large-scale panoramic dataset for outdoor scene reconstruction. Extensive experiments demonstrate that G$^2$PS achieves state-of-the-art rendering quality while maintaining scalable, block-parallel training. Our models, training code, and dataset are publicly available.

The rapid development of large language models and multimodal large language models has accelerated the emergence of proactive agents capable of operating everyday tools and assisting users in real-world environments. However, existing benchmarks struggle to evaluate such agents effectively, as they often rely on sandboxed environments and single-turn evaluation paradigms. Moreover, their scenario-based task taxonomies mix multiple model capabilities within the same task category, making it difficult to identify the root causes of agent failures. To address these limitations, we introduce UniClawBench, the first capability-driven benchmark designed to evaluate proactive agents in dynamic, real-world settings. UniClawBench is built around five foundational model capabilities: Skill Usage, Exploration, Long-Context Reasoning, Multimodal Understanding, and Cross-Platform Coordination. Based on these capabilities, we design 400 bilingual real-world tasks. Unlike previous benchmarks that rely on static, pre-recorded answers, our benchmark evaluates agents in live Docker containers using fine-grained, step-by-step completion checkpoints. Furthermore, we design a closed-loop evaluation strategy comprising an executor agent, a hidden supervisor agent, and a user agent to simulate realistic multi-turn human feedback without leaking grading criteria. To disentangle base model capabilities from framework-level design choices, we evaluate state-of-the-art models under multiple agent frameworks. Through comprehensive comparisons across both models and frameworks, we show how base model capabilities and agent framework designs jointly shape performance in real-world environments. To facilitate future research, we make our benchmark and code publicly available at https://github.com/HKU-MMLab/UniClawBench.

We propose OPSD-V, an on-policy self-distillation paradigm for post-training few-step autoregressive (AR) video diffusion models. Existing few-step AR video generators can produce long videos with low latency, but still suffer from error accumulation and weakened motion dynamics during long autoregressive rollout. OPSD-V reduces long-horizon degradation while preserving the original few-step inference path. The key idea is to introduce real long-video data as temporal context during training and use it to provide dense trajectory-level supervision. Specifically, the student follows the exact inference-time rollout, generating each chunk conditioned on its own previously generated KV cache. In parallel, the teacher is evaluated at the same student-visited denoising states, but uses a cleaner AR-consistent temporal cache in which older history can be replaced by real-video context. This provides dense denoising-level corrective targets under on-policy AR cache dynamics, without changing the sampler, number of denoising steps, or inference-time cache mechanism. We apply OPSD-V to representative few-step AR video models, including Self-Forcing and LongLive. Experiments show consistent improvements in visual quality, motion dynamics, and VBenchLong scores. A user study with 10 participants comparing 20 video pairs shows that OPSD-V is preferred over the base models in 66.0% of overall-preference judgments (82.5% excluding ties).

Reasoning has become a core capability for large models, especially when reliable decisions require understanding logical consequences. Recent video generation models offer a reasoning path distinct from previous Chain-of-Thought (CoT): reasoning can unfold through temporally connected frames, known as Chain-of-Frame (CoF) reasoning. However, existing video generators are primarily trained on general video corpora, still lacking diverse supervision and dedicated designs for CoF reasoning. To address this gap, we introduce OpenCoF, a framework comprising the OpenCoF-17K dataset, a reasoning video dataset spanning 11 task families, and Wan-CoF, a fine-tuned video model for studying whether diverse temporal supervision improves CoF behavior. Across four video reasoning benchmarks, Wan-CoF achieves considerable gains over the Wan2.2-I2V-A14B baseline. Building on this, we empirically explore more advanced designs for CoF capabilities, i.e., equipping the model with visual and textual reasoning tokens. This mechanism respectively captures low-level visual cues and high-level semantic priors for spatial and temporal reasoning. Through performance comparisons and attention analysis, we examine how these tokens contribute across model depth, denoising steps, space, and time. Our results suggest that stronger video reasoning requires both broad temporal supervision and explicit mechanisms for organizing intermediate reasoning state. We open-source the dataset, model, and code to facilitate future research on reasoning-oriented video generation.

Scientific ideas rarely start from a blank page. They inherit mechanisms, repair known limitations, and recombine pieces of earlier work, much like biological genomes. Current benchmarks still say little about whether AI systems can follow this inheritance structure. We present IdeaGene-Bench (IG-Bench), a benchmark for scientific lineage reasoning and lineage-grounded idea generation. IG-Bench is organized around the IdeaGene framework: each paper or proposal is represented as a set of minimal, typed, evidence-grounded Idea Genome objects, and a GenomeDiff aligns these objects to record inheritance, mutation, loss, external import, and novel insertion under six operational evolutionary dynamics. The benchmark contains 1,961 golden lineage traces, 1,085 curated Idea Genome objects, and 920 pairwise GenomeDiff records across 10 scientific domains. It supports two evaluations. IG-Exam (42 task types, 1,029 instances) tests closed-form lineage reasoning across Idea Genome abstraction, inheritance tracing, evolutionary reasoning, and lineage verification. IG-Arena evaluates generation with a lineage-conditioned Population-Evolution Score(PES), asking whether a proposal can be inserted as a coherent descendant of a given lineage population: it should inherit the right Idea Genome objects, vary meaningfully from nearby work, and offer selection value for future research. Experiments on 14 LLM-based scientists expose a compositional bottleneck. The strongest system reaches only 27.3% exact accuracy on lineage reasoning, and structured lineage context reshuffles system rankings rather than helping every participant uniformly.

Building general-purpose dexterous manipulation policies requires benchmarks that go beyond isolated tasks to systematically evaluate policies across diverse interaction modes, sensory conditions, and robot embodiments. However, existing benchmarks remain limited in task and data diversity, embodiment coverage, or controllable visual variation, hindering studies of cross-task and cross-embodiment generalization. We present DexVerse, a large-scale and modular benchmark for dexterous manipulation. DexVerse includes 100 tasks spanning a broad range of manipulation skills, including object grasping and relocation, articulated-object interaction, functional tool use, bimanual coordination, non-prehensile control, contact-rich behaviors, multi-goal execution, and long-horizon multi-stage task completion. It supports 3 robot arms and 6 dexterous hands, and is extensible to new tasks, assets, and embodiments. To evaluate visuomotor generalization, DexVerse provides configurable visual variations in textures, background, lighting, and camera viewpoints. We further provide a VR-based teleoperation interface and 3,180 demonstrations with synchronized proprioceptive, RGB, depth, point-cloud, and state observations. We benchmark representative methods, including Diffusion Policy, DP3, OpenVLA, and $π_{0.5}$, across 19 tasks. Results reveal substantial challenges in task generalization and visuomotor robustness, establishing DexVerse as a promising testbed for general-purpose dexterous manipulation. Project page: https://ycyao216.github.io/DexVerse.site

Recent advances in Vision-Language Models, Large Language Models, and Multimodal Large Language Models have improved autonomous driving tasks such as scene understanding, decision making, trajectory prediction, and visual question answering. However, evaluating whether these models can reliably reason about safety-critical incidents remains challenging. To address this gap, we present AUTOPILOT-VQA, an incident-centric visual question answering benchmark for dashcam video understanding. The dataset evaluates different systems through structured questions designed around real-world driving incidents and near-incidents. The benchmark covers diverse safety-relevant categories, including weather and lighting conditions, traffic environment, road layout, road surface state, signage, involved entities, accident occurrence, impact location, and avoidability-related reasoning. By requiring models to answer grounded questions about both contextual scene properties and event-level incident details, AUTOPILOT-VQA moves beyond object recognition toward temporally grounded, safety-aware reasoning. The dataset is released as part of the AUTOPILOT CVPR 2026 competition and provides a standardized benchmark for assessing the reliability of autonomous driving systems in different scenarios. Our benchmark support developments for more interpretable, robust, and safety-conscious vision-language systems for real-world autonomous driving.

Keypoint tracking alone is insufficient for object interaction tasks such as sitting on a chair, wiping a board, or pushing furniture, where the robot can reach the correct pose without making meaningful physical contact with the object. We present CONTACTMIMIC, a learning framework that tracks explicit partlevel binary contact commands alongside keypoint trajectories. CONTACTMIMIC is made possible through the use of contact-following rewards and a trajectory augmentation scheme aimed at breaking the correlations between keypoint trajectories and contact labels. The resulting policy successfully decouples contact behavior from keypoint geometry, and achieves precise physical contact as well as contact-controllability (produce or suppress contact during deployment as desired). Simulation experiments across 10 diverse human-object interaction motions confirm that CONTACTMIMIC exhibits contact controllability that enables it to complete manipulation tasks without task-specific rewards, while also outperforming keypoint-only trackers on contact-relevant tasks. Ablations confirm the necessity of the proposed trajectory augmentation scheme and sim2real deployment validates contact controllability in the real world across 5 different motions. Video results are available on https://lixinyao11.github.io/contactmimic-page/.

Generating realistic 3D human motions in real-time within interactive applications is key for animation, simulation, and humanoid robotics. While recent offline motion generation approaches offer precise control via text and kinematic constraints, they lack the inference speed required for interactive settings. Conversely, existing online methods enable real-time synthesis but often sacrifice controllability or struggle with complex text semantics and long-horizon goals due to limited context windows. In this work, we introduce ARDY, a streaming generation framework that bridges this gap by enabling high-fidelity motion generation controllable via online text prompts and flexible kinematic constraints. ARDY employs a hybrid representation that combines explicit root features with a latent body embedding, balancing precise trajectory control with efficient generative learning. We propose a two-stage autoregressive transformer denoiser that features variable history context and supports conditioning on flexible, long-horizon kinematic constraints. By training on a large-scale motion capture dataset and being directly conditioned on text labels and kinematic constraints sampled from ground truth poses, ARDY natively learns controllable generation that supports online prompting and flexible long-horizon goals. Extensive evaluations on the HumanML3D benchmark and the large-scale, high-fidelity Bones Rigplay dataset demonstrate ARDY's high motion quality and constraint adherence, validating the efficacy of our key architectural decisions. Finally, we demonstrate the method's practical versatility through an interactive demo featuring dynamic text control, diverse keyframe pose constraints, path following, and interactive locomotion control via mouse and keyboard. Supplementary video results, code, and model releases can be found at https://research.nvidia.com/labs/sil/projects/ardy/.

Modern robotic perception increasingly involves large-scale geometric optimization problems distributed across multiple robots or sessions. However, existing distributed solvers often depend on brittle hand tuning and primarily target rigid body pose graphs. To address this, we present DeepCORD, a learning-augmented framework for distributed factor graph optimization on general matrix Lie groups. By unfolding a parallel and accelerated Riemannian optimizer into differentiable iterations, DeepCORD learns a self-supervised feedback policy that dynamically adapts solver parameters according to the optimization phase and communication status. The resulting method enables adaptive distributed optimization over matrix Lie groups under both synchronous and asynchronous communication regimes. Extensive experiments on real-world $\mathrm{SE}$(3) pose graph optimization and $\mathrm{SL}$(4) projective submap alignment show that our method achieves lower objective values than existing distributed baselines on most benchmarks across realistic operating scenarios.

A national language model offers a linguistic community its own instrument for measuring what its citizens say and value. Portugal's AMALIA, a publicly funded 9B-parameter model for European Portuguese, appears competitive on agreement alone: asked to code the moral foundation of authority, it agrees with trained human coders to within six F1 points of open models eight to thirteen times its size. Yet agreement is reliability, not validity. For theoretical constructs that must be inferred rather than read from surface features, the question is whether the model follows the construct's theory or reaches the right code by correlated shortcuts. We test this with the recovery gap: the loss in performance when a holistic prompt is decomposed into the codebook's atomic clauses and recombined by the theory's explicit rule. If calibration closes that gap, some portability should survive across models and languages; where it does not, the construct-model instrument is the likely locus of failure. We ask whether a calibrated English instrument transfers to AMALIA-9B and to European Portuguese. For one construct and one corpus, it does not. Decomposition recovers only about half of AMALIA's holistic performance, and error analysis suggests reliance on surface correlates, especially moral outrage near authority figures. An open multilingual LLM closes the gap on the same Portuguese corpus under the same instructions, pointing away from the corpus as the main explanation. AMALIA can still screen and pre-code at scale, but it cannot yet measure this construct well enough to stand alone. The study is a single counterexample, not a verdict on national models; it argues that sovereign-LLM benchmark batteries should test not only agreement with human coders, but the evidential route by which that agreement is warranted.

Multi-object tracking (MOT) has achieved strong performance on benchmarks dominated by short video sequences. However, such datasets do not adequately evaluate long-term identity preservation, where objects must be tracked consistently over extended durations. We introduce WaspMOT, a benchmark designed to address this gap through long-duration tracking of Trichogramma wasps in controlled ecological experiments. The dataset contains 10 sequences of approximately 12,000 frames each (over 8 minutes at 25 FPS), with dense MOTChallenge annotations and oracle detections to isolate association performance. Unlike existing benchmarks, WaspMOT forms a closed-set tracking scenario where all individuals remain present throughout the sequence, requiring consistent identity assignment across thousands of frames despite abrupt jumps, occlusions, and highly similar appearance. We establish a benchmark by evaluating five tracking-by-detection methods, including ByteTrack, BoT-SORT, C-BIoU, OC-SORT, and McByte, under a unified protocol. Results show that all methods suffer from significant trajectory fragmentation, highlighting the difficulty of long-term identity preservation even with perfect detections. A simple spatial tracklet stitching baseline consistently improves performance, indicating that substantial gains remain possible. WaspMOT provides a new benchmark for studying long-term association and reveals limitations of current tracking approaches that are not observable on conventional datasets. The benchmark will be made publicly available at the project repository: https://github.com/tstanczyk95/WaspMOT/ .

Recent progress in 3D human pose estimation has made markerless recovery of skeletal motion increasingly accurate and scalable. However, most pose estimators remain optimized for geometric keypoint accuracy, while many real-world applications in rehabilitation, sports science, ergonomics, and clinical movement analysis require biomechanical quantities that describe how the body moves, loads, and activates. In this work, we propose BioModule, a lightweight plug-in temporal transformer that attaches downstream of any 3D pose estimator and predicts biomechanical attributes from standard 17-joint 3D skeletons. BioModule is estimator-agnostic and requires no modification of the upstream pose model, enabling existing pose estimators to be extended toward physically interpretable motion analysis. To train and evaluate BioModule, we construct a large-scale aligned dataset pairing Human3.6M video and 3D keypoints with the biomechanical label space of Human3.6Mplus. We establish and verify anatomical correspondence between coordinate systems of the two datasets, enabling frame-accurate cross-modal supervision. Using this aligned supervision, BioModule predicts biomechanical quantities. We further benchmark BioModule across seven state-of-the-art 3D pose estimators, providing the first systematic analysis of how upstream pose estimation quality propagates to downstream biomechanical prediction fidelity. The results position BioModule as a compact, modular bridge between vision-based pose estimation and biomechanically meaningful human motion analysis.

Human decision-making is highly flexible -- some actions are taken immediately; others require longer deliberation. Language models have exhibited a similar capacity for adaptive "reasoning." However, transferring this capability to continuous control policies has been challenging, as directly reasoning in language space may lack the granularity for spatial understanding and precise motions. In this work, we show that reasoning for control policies can emerge by organizing information in an autoregressive latent space reminiscent of a memory palace, where retrieval is iterative and adaptive. Our method, Latent Memory Palace (LMP), formulates reasoning as variational inference with an autoregressive latent distribution. We derive a latent-space reinforcement learning technique to tractably optimize its variational lower bound. The resulting policy, LMP-$π$, achieves strong empirical performance in simulation and real-world domains while exhibiting interpretable, adaptive allocation of test-time compute. We further show that the same framework yields a variable-length action tokenizer, LMP-$\texttt{tok}$, which significantly improves the performance of downstream autoregressive policies. Together, these results present a new perspective on latent reasoning for control through the lens of variational inference.

In long-horizon tasks, decision-relevant state is often scattered across an expanding trajectory, while the action agent must surface it and act. As trajectories grow, task requirements, environment facts, prior attempts, diagnoses, and open subgoals can be buried in the context window or pushed beyond it, failing to influence decisions when needed. We call this failure mode "behavioral state decay". We study memory as an active intervention mechanism rather than passive retrieval. A separate memory agent runs alongside an unmodified action agent, updating a structured memory bank from the recent trajectory and deciding whether to inject a memory-grounded reminder or remain silent. The module is plug-and-play with frontier action agents and existing agent harnesses. Across Terminal-Bench 2.0 and $τ^2$-Bench, it improves pass@1 for both weaker and stronger action agents, with gains of +8.3 pp on Terminal-Bench and +6.8 pp on $τ^2$-Bench. Ablations show that selective intervention outperforms passive bank exposure, always-on injection, advisor-only guidance, and general retrieval. As an early step toward open-weight memory policies, we train Qwen3.5-27B on SETA using SFT and GRPO, improving validation reward and achieving partial transfer to Terminal-Bench.

Rapid advancements in video diffusion models and temporal editing tools have enabled the generation of highly realistic human-centric videos, posing unprecedented challenges to digital content forensics. Existing benchmarks primarily focus on either face-swapping or global text-to-video synthesis, overlooking the crucial dimensions of human-object or human-human interactions and multi-modal alignment. To address these limitations, we introduce HumanForge, a unified, large-scale, and multi-paradigm human-centric video forgery dataset. To construct and annotate this dataset without labor-intensive manual labeling or hallucinated monolithic prompts, we propose Gen2Anno, a modular active multi-agent pipeline built on LangGraph. Gen2Anno coordinates six specialized agents-ranging from source profiling to MoE-based reference analysis and closed-loop forensic verification-to generate over 18K high-fidelity video segments and produce structured, contrastive omni-annotations containing binary decisions, fine-grained artifact categories, and spatio-temporal localization. Extensive benchmarks using state-of-the-art traditional detectors and Large Multimodal Models (LMMs) demonstrate the significant challenges of zero-shot generalization and fine-grained reasoning on HumanForge. Code and dataset will be publicly released.

Reinforcement learning increasingly relies on an LLM judge to score each rubric criterion, and that judge acts as the reward model during training. Before such a signal can be trusted, we need to know how capable the judge must be and how biased it is. We study this calibration question for citation quality in deep-research systems, where a search-grounded LLM must support each claim it writes with a cited source. Citation quality is a structured rubric task in which each attribution-citation pair is judged along two dimensions that require an LLM, source relevance and factual support. On an adversarial long-form benchmark, we score 8 off-the-shelf LLM judges from 3 model families against gold labels over 1,248 rubric decisions, all of which were human-reviewed and 378 of which were hard cases adjudicated from judge disagreements. Cheaper judges remain competitive across both dimensions, with GPT-5-mini attaining the strongest source-relevance pass-class F1 at 0.908 ($κ$=0.636), while on factual support the judges are statistically indistinguishable (overlapping confidence intervals), so no single model dominates. At comparable F1, the judges still differ substantially in pass-rate drift, false positive rate, and false negative rate. Scalar F1 obscures this directional bias, yet it is exactly what a downstream reinforcement learning loop would reinforce. Calibrating the judge is therefore a prerequisite for using citation rubrics as reward signals, and our results show that this calibration does not require the most expensive available model.

Modern Video Object Segmentation (VOS) involves tracking and segmenting user-specified targets. While recent approaches have achieved remarkable performance in single-target scenarios, extending them to multi-target settings typically involves replicating the single-target processing for each individual object, resulting in reduced frame rates (FPS) with unbounded latency as target count increases. Built upon Segment Anything 2 (SAM2), we propose SAM-MT, which addresses this by transforming the model into an interactive framework for real-time Multi-Target video segmentation. SAM-MT uses explicit queries to represent different individual targets, in parallel with a shared representation for global context. It employs decoupled masked attention to keep individual identities distinct from cross-target interference, and sparse memory for stable temporal evolution, along with specialized strategies for occlusion handling and overlap prevention. SAM-MT successfully decouples latency from the number of targets, achieving real-time speed on par with single-target baselines (>36 FPS for 10 targets) while maintaining SAM2's robust video segmentation performance.

As agentic AI systems are increasingly applied to cyber-physical environments, their evaluation requires assessment of both task performance and trustworthiness. In decentralized energy markets, autonomous agents may improve market utility, but may also exploit invalid physical data, create artificial liquidity, and produce unstable governance decisions. Therefore, we propose SolarChain-Eval, a physics-constrained benchmark for evaluating trustworthy economic agents. It formulates market governance as a Gymnasium-compatible Markov Decision Process, where agents make hourly decisions. SolarChain-Eval evaluates each policy across multiple dimensions, including market utility, physical safety, slippage, action smoothness, spatial fairness, and auditability. To support agentic evaluation, SolarChain-Eval incorporates an LLM-based Planner/Auditor layer. The Planner defines episode-level action bounds and audit rules, while the Auditor reviews and revises high-risk actions. All interventions are recorded through structured logs, including trigger signals, proposed actions, revised actions, and audit rationales. Experiments with static, random, myopic, RL, and RL+LLM policies reveal a clear utility-safety trade-off. RL agents improve market utility but can still produce unsafe behavior. When the physics penalty is removed, reward-maximizing agents exploit invalid generation and increase artificial liquidity. The LLM Planner/Auditor improves auditability and mitigates selected risks, but it cannot fully compensate for a misspecified reward function. These results indicate that trustworthy agentic AI evaluation requires both physical constraints and transparent intervention traces. We release data and code as open access on GitHub for replicability.

The rapid advancement of generative AI has enabled the creation of highly realistic deepfake media, posing significant threats, including misinformation, digital identity theft, fraud, and manipulation of public opinion. AI-generated image (AIGI) detection is reliably challenging due to the diversity of generative methods and the subtle artifacts they leave behind. In this work, we propose GenRes, a novel framework for generative residual learning via a neural tensor network, which models fine-grained relational features between original and transformed samples to enhance generalization. To address scenarios involving multiple generative transformations, we introduce GenRes++, which employs a learnable attention mechanism to aggregate relational features across multiple transformed samples and enables the model to focus on the most informative cues. Both models leverage PE-Core as a feature extractor, providing generalized and semantically rich embeddings that improve cross-domain performance and enable the detection of AIGI generated by unseen methods. Comprehensive experiments on multiple benchmark datasets demonstrate that the proposed GenRes++ approach outperforms existing methods.

Large language model (LLM)-based web search agents are transforming information seeking from simple factoid question answering into complex, deep-and-wide search and research-oriented tasks. A single ReAct-style agent is constrained by one long trajectory and limited context, making it difficult to handle depth and coverage simultaneously. Existing multi-agent systems improve search coverage through parallel execution and aggregation, but still exhibit clear limitations in recursive depth, collaboration adaptability, and evidence-grounded expansion. We propose WebSwarm, a progressive recursive delegation framework that jointly constructs task decomposition, recursive expansion, and agent collaboration during inference. WebSwarm dynamically instantiates agentic search nodes, each coupling a local objective with a search mode that specifies how the node should organize search and collaboration. Each node can either solve its objective itself or further delegate child nodes; after solving, it returns evidence and results upward, enabling parent nodes to further expand, revise, or aggregate the search process. To guide this process, WebSwarm first probes how task-relevant information is organized on the web to ground subsequent node expansion, and reuses process-level experience across homogeneous sibling nodes. Experiments on BrowseComp-Plus, WideSearch, DeepWideSearch, and GISA show that WebSwarm consistently outperforms single-agent and multi-agent baselines on deep, wide, and interleaved deep-and-wide tasks. Further analyses of ablation, task difficulty, web tool efficiency, and model generalization explain WebSwarm's effectiveness and provide insights for multi-agent search systems.

Speculative decoding accelerates LLM inference by drafting several tokens and verifying them in parallel. Block-diffusion drafters such as DFlash produce a draft block in one pass but model only per-position marginals, and best-first tree methods such as DDTree expand candidate trees from those marginals. The released Domino drafter adds a GRU-based causal correction that makes each draft token distribution path-dependent, a structure DDTree's factorized formulation cannot represent. We introduce DominoTree, a training-free best-first draft tree scored by Domino's conditional, non-factorized correction along each root-to-node path, made practical by restricting the per-node correction to a candidate top-M set. On Qwen3-4B across eight benchmarks, DominoTree reaches up to 6.6x speedup over autoregressive decoding and the highest mean accepted length of any evaluated method, up to 10.7 tokens per round, at every tested temperature. DominoTree constructs its tree with a GPU-native CUDA-graph builder that is bit-identical to a reference Python implementation, so acceptance is unchanged, while keeping per-round tree construction cheap. With this builder as default, DominoTree improves throughput over the released Domino decoder, the drafter it builds on, at every tested temperature: 9% to 10% overall on Qwen3-4B and up to 22% on Alpaca. It also outperforms DDTree and CaDDTree at every tested temperature, not only under greedy decoding. On Qwen3-8B, DominoTree keeps the highest accepted length at every temperature and gives a 24% throughput gain over DDTree at T=0; at higher temperature its edge over DDTree and CaDDTree narrows to a tie and a small loss, while its aggregate gains over DFlash and Domino persist.

The advent of video-action models offers a promising path for robot control. Nevertheless, we argue that repurposing video generative models designed for digital content creation is inherently inadequate for physical environments. To bridge this gap, we present LingBot-VA 2.0, a video-action foundation model built from the ground up for embodiment. Four core design principles showcase its evolution from LingBot-VA. (1) Departing from traditional reconstruction-focused VAEs, we introduce a semantic visual-action tokenizer, which aligns visual representations with both semantics and actions, improving instruction following and action precision in subsequent policy learning. (2) Given the strictly causal nature of temporal dynamics, we adopt a causal pretraining paradigm, training from scratch to circumvent the catastrophic forgetting that frequently occurs when adapting bidirectional architectures. (3) To meet the demands of high-frequency inference, our model employs a sparse MoE backbone, expanding model capacity without compromising efficiency. (4) Real-time closed-loop control is realized through an enhanced asynchronous inference scheme, which predicts future latents in parallel with action execution while re-grounding each rollout on the latest observation via learned forward dynamics. Real-world deployment validates LingBot-VA 2.0 as a robust foundation model, as evidenced by its few-shot generalization across complex manipulation tasks.

As human-robot interaction rapidly spreads in numerous fields, the subject of robot acceptance gains increasing importance. Visual similarity to the human body, as occurs for humanoids, is generally not enough to ensure acceptance in physical interaction, as acceptance directly links to comfort and ergonomics, which are measured in terms of the quality of the robot movement perceived by the human. This paper discusses the connection between comfort and similarity of the robot movement to the human one. By considering the kinematic characterization of human movement, this paper focuses on the time laws of such movements, wherein the end-effector path is prescribed. Based on the lognormality principle for modeling human movements, a human-likeness index is defined and used to provide an a priori characterization of trajectories. Such an index can be used to evaluate the performance of trajectory generation algorithms in producing human-like movements before they are actually executed. For validation purposes, 68 subjects are required to judge their comfort. The results of three experimental campaigns involving a physical interaction with a robot demonstrate a globally consistent trend between the preference in terms of perceived comfort and the distribution of the suggested human-likeness index.

Sparsely-activated Mixture-of-Experts (MoE) language models achieve remarkable inference efficiency by activating only a small fraction of parameters per token, yet their full expert banks reside in memory at all times, creating a prohibitive deployment bottleneck. Existing structured pruning methods, largely designed for dense transformers, assess expert importance using locally derived heuristics that are blind to the interdependent nature of MoE routing. We introduce MAESTRO (Markov-chain Approximated Expert Sparsification via Transition-based ROuting), a structured pruning framework designed for MoE architectures that models autoregressive expert activation trajectories as Ergodic Markov chains whose stationary distributions encode cross-layer dependencies, yielding a globally aware importance heuristic. Evaluated across five diverse domains including Safety, Bias, and Ethics, MAESTRO outperforms state-of-the-art baselines by up to 10.61% in average performance retention under a strict 50% compression regime, while exhibiting substantially lower cross-task variance, indicating that global, routing-congruent pruning produces models that generalize more consistently across heterogeneous tasks.

We present FabriVLA, a lightweight Vision-Language-Action model for Precise Multi-Task Manipulation. FabriVLA combines an InternVL3.5 vision-language backbone with a flow-matching action head featuring gated self-attention across action tokens and shallow VLM layer fusion for enriched spatial context. The model is trained via single stage joint optimization from a pretrained VLM and randomly initialized action head. On the Meta-World MT50 benchmark spanning 50 diverse manipulation tasks, FabriVLA achieves a tier-average success rate of 90.0%, demonstrating that a compact VLA built on a 1B scale VLM can achieve strong performance without relying on multi billion parameter VLA backbones.

Multimodal Large Language Models (MLLMs) often follow a fixed Think-then-Answer paradigm, which is inefficient in heterogeneous multitask settings because simple inputs may not require explicit reasoning while difficult ones can benefit substantially from it. Learning when to think is also unstable during post-training, where imbalanced rollouts can drive the model toward always-thinking or always-direct behavior. We propose Switch-Reasoner, a GRPO-based framework that learns to adaptively select reasoning modes for MLLMs. It treats thinking as a virtual tool invocation and allows the model to either answer directly or invoke explicit reasoning before answering. To stabilize this decision, we introduce a dual-level regulation mechanism that balances the overall use of Thinking Mode and Direct Mode while providing sample-level supervision based on the relative benefit of the two choices. Experiments on 11 multimodal tasks show that Switch-Reasoner reduces unnecessary reasoning while maintaining strong performance, achieving a better accuracy-efficiency trade-off.

While initially deployed for optical interconnects, silicon photonics is increasingly being explored as a hardware platform for programmable optical systems, including linear optical processors, neuromorphic photonic networks, quantum photonic circuits and multiplexed sensor arrays. Common to most existing implementations is that light is controlled with electronics, and even basic demonstrations wherein light directly controls light remain limited. Here we demonstrate a broadband all-optical silicon photonic phase shifter based on an optomechanically mediated light-light interaction arising from the gradient optical force. Our device concept relies on slot-mode waveguides suspended by subwavelength gratings, which provide mechanical support while preserving optical confinement. We demonstrate all-optical phase shifting using a guided pump beam co-propagating with the signal beam, with only 60 $μ$W required to achieve a $π$ phase shift in a 178.6 $μ$m-long device. In addition, we measure the required pump power across a wide parameter space and find quantitative agreement with a lumped force-equilibrium model. Since the actuation relies on an all-optical geometric deformation rather than on material-index tuning, the approach avoids local electrical connections to the active element, carries no Kramers-Kronig absorption penalty, and is naturally compatible with cryogenic quantum photonic platforms.

The recently established 'Out of Sight, Not out of Mind' (OSNOM) task for egocentric videos focuses on tracking objects that are moved by the camera wearer, online, maintaining knowledge of instance locations throughout the video even when they leave the field of view or become heavily occluded. In this paper, we propose the first learning-based solution to the OSNOM task: Whareformer, a transformer-based model with two components: an updatable memory of established tracks and a track assignment module that associates observations with existing tracks in a feed-forward manner. Whareformer jointly reasons over evolving object appearance (what) and updated 3D location (where), and employs a dedicated New Track token to reason about novel objects. Thanks to its design choices of using relative distances and evolving track representations, Whareformer is trained on a small set of 56 videos but achieves SOTA performance on 260 long test videos from three datasets: EPIC-KITCHENS-100 (unseen videos), IT3DEgo, and HD-EPIC, with significant absolute improvements over prior work.

An LLM-as-judge score can move even when the candidate responses stay fixed, simply because the evaluator has changed. We treat this evaluator-replacement ambiguity as a measurement-validity problem. Across four judgment datasets, we compare two upgrade paths available in practice: scaling Qwen3 dense judges from 1.7B to 32B parameters and moving across MiniMax M2-M2.7 released APIs. The main pattern is that judge upgrades are not interchangeable: only Qwen3 1.7B to 4B gives a robust adjacent gain, while MiniMax adjacent releases do not. Stronger judges reduce but do not remove position and verbosity bias. Repeated-sample juries add little when errors are correlated. Structured debate can move decisions substantially, but without parser and fallback logs those shifts cannot be attributed to deliberation. We argue that LLM-as-judge reports should include dataset slices, bias probes, error-dependence estimates, and protocol audit trails.

Text-to-image (T2I) models have been shown to exhibit social biases. Prior work has mainly focused on gender, skin tone, and cultural representation within restricted occupational associations, and emerging benchmarks increasingly incorporate these dimensions. However, disability remains systematically underexplored. Current evaluation practices often fail to align with sociologically grounded definitions of stereotyping, limiting principled assessment of representational harms toward people with disabilities (PWD). To address this, we introduce INCLUDE-BENCH, the first large-scale benchmark for evaluating disability-related bias in T2I models. INCLUDE-BENCH comprises 119K generated images based on prompt design across multiple bias dimensions and both static and dynamic contexts. We evaluate 15 open-source and 2 closed-source models. Our key findings reveal that: (1) mobility-impaired and default disability prompts predominantly yield wheelchair depictions across all models; (2) disability-conditioned generations consistently exhibit less diversity; (3) stereotypical portrayals demonstrate stronger disability-text alignment; and (4) we introduce the Stereotype Content Model (SCM) Score, demonstrating that T2I models reflect real-world stereotypical associations.

Hand-object interaction (HOI) recognition requires capturing both hand manipulations and object transformations. However, existing video-language models often fall into shortcuts by relying on spurious correlations among hands, objects, or environmental context, rather than reasoning from the appearance and dynamics of hands and objects themselves. To address this limitation, we propose a new learning paradigm that combines (i) hand-object masked training, which enables robust reasoning from partial hand or object observations, and (ii) an HOI-dynamics-aware decoder that explicitly learns hand- and object-centric embeddings through auxiliary predictions of their locations and semantics, enhancing sensitivity to both cues. To systematically evaluate such cue-specific reasoning, we introduce Cue-Isolated HOI (CI-HOI), a new evaluation that assesses models' ability to predict actions from hand- and object-related cues independently. To enable CI-HOI, we curate the DEHOI testbed, which separates hand- and object-related observations for disentangled HOI evaluation through inpainting. Using DEHOI, we demonstrate both quantitatively and qualitatively that our training strategy exploits hand- and object-centric information more effectively than existing models. Our approach improves over existing models on DEHOI, standard action recognition, object state recognition, and even robot manipulation action recognition, leading to more robust HOI understanding.

Most cooperative Vision-Language Navigation (VLN) methods assume unlimited communication, not considering real-world applications where bandwidth is restricted and information efficiency is critical. We introduce \textbf{bandwidth-constrained cooperative VLN} and propose \textbf{hindsight gating}: a lightweight supervised gate that labels communication-critical steps post-hoc from navigation failures, avoiding the high variance of REINFORCE. Contrary to the intuition that agents should communicate when uncertain, we observe a consistent counter-intuitive pattern: trained gates fire predominantly in early episode steps and more often when agents are confident, across all budget levels ($B \in \{1,3,5\}$). We explain this through \textbf{recurrent hidden-state alignment}: early communication injects grounded trajectory representations that persist and compound through subsequent Gated Recurrent Unit (GRU) updates, achieving $+0.072$ cumulative alignment gain with $B{=}3$ transmissions, approaching unconstrained communication ($+0.078$) at 260\% greater alignment efficiency than random gating ($+0.020$) and 320\% greater efficiency than entropy-based gating ($+0.017$). Our results establish a new communication regime for bandwidth-limited embodied agents: synchronise representations early, navigate independently later. Our codebase is available at: https://github.com/AravG13/bandwidth-constrained-cooperative-vln

We present a Procrustes-conditioned Joint End-to-end Top-K Sparse Autoencoder (SAE) for extracting cross-seed universal features from independently trained BERT models. Cross-seed feature universality is a fundamental challenge in mechanistic interpretability: because dictionary learning is non-convex, independently trained networks learn misaligned feature spaces, so apparently identical features may differ by random initialization. We address this by computing an orthogonal Procrustes rotation between seeds' activation spaces before joint SAE training, combining Top-K sparsity, end-to-end downstream optimization, and an auxiliary dead-feature revival loss based on previous SAE literature. Evaluating on five independent seed pairs (ten BERT models) across three benchmark datasets (SST-2, Stanford Politeness, TweetEval Emotion), our full pipeline produces more universal features (Pearson r $\geq$ 0.70 across seeds) than post-hoc alignment baselines on all three datasets. A minimal qualitative analysis confirms that high-universality features encode interpretable sociolinguistic patterns.

Recent unified multimodal models show a single architecture can jointly perform vision/language understanding and image generation/editing. However, they repeatedly feed all historical visual and textual inputs into a shared context window, limiting long-horizon multimodal dialogue due to visual token explosion and unreliable cross-turn referencing. We propose a Cognitive-structured Multimodal Agent that externalizes visual information into an Episodic Visual Memory and selectively reactivates relevant episodes during reasoning. The agent consists of a Perceptual Abstraction Engine for structured visual abstraction, a Cognitive Retrieval Engine for cross-turn memory retrieval, and a Multimodal Executive Controller for autonomous task inference and action planning. To address the lack of turn-level retrieval supervision in existing datasets, we develop a Unified Scenario Engine that programmatically generates structured multi-turn conversations with fine-grained retrieval annotations, enabling reinforcement learning to optimize abstraction and retrieval policies. We also construct a long-horizon visual-dialogue benchmark stratified by difficulty to evaluate episodic visual recall. Our 8B agent achieves 91.4% retrieval accuracy over 20-turn sessions, surpassing 32B baselines by +8.2% while nearly halving per-turn inference time (23.1s -> 12.7s). We further present the Cognitive-structured Multimodal Agent Harness (CMA-Harness), a tool-augmented deployment of the same cognitive structure integrating persistent multimodal memory, web access, image generation/editing/composition tools, and OpenAI-compatible serving. Structured memory and modular decision-making offer a more scalable, efficient paradigm for long-horizon multimodal agents than monolithic parameter scaling. Code: https://github.com/caseclose/cma-harness ; Project page: https://caseclose.github.io/cma-harness/

Subjective NLP tasks often exhibit systematic annotator disagreement, requiring models that represent uncertainty rather than collapse it. We introduce Ensemble Diversity Optimization (EDO), a prediction-space framework that jointly optimizes ensemble weights, effective cardinality, and calibration through a unified differentiable objective. EDO learns ensemble composition and size end-to-end via Gumbel-Softmax relaxation and incorporates a signed diversity regularizer, tuned on validation data, to steer optimization toward either preserving or suppressing disagreement. This regularization prevents ensemble collapse and enables controlled navigation of the utility-calibration trade-off. The framework integrates a soft F1 surrogate, class-weighted cross-entropy to address imbalance, and reliability-weighted diversity to regulate intra-ensemble variability. Experiments on four subjective text-classification benchmarks (ArMIS, ConvAbuse, HS-Brexit, MD-Agreement) show that EDO substantially improves probabilistic calibration, reducing cross-entropy (40-78% depending on baseline) and lowering Brier scores relative to Soft-CE, Soft-MD, Top-5 Voting, and WEL, while maintaining competitive F1 and better alignment with annotator distributions. These results demonstrate that jointly optimizing ensemble structure with a signed diversity regularizer provides an efficient, model-agnostic approach for modeling human subjectivity in supervised learning.

Language-conditioned manipulation requires both precise contact-rich control and robust reasoning over language, scenes, and long horizons. End-to-end Vision-Language-Action (VLA) models provide strong local visuomotor skills, but they are trained on in-distribution task trajectories and often fail under deployment perturbations such as semantic retargeting, goal re-binding, spatial-layout shifts, and unstable local contacts. LLM coding agents provide complementary semantic and compositional reasoning, but purely analytic primitives struggle with irregular grasping, constrained placement, and articulated-object interaction. We present Harness VLA, a memory-augmented agentic framework that exposes a frozen VLA as a retryable contact-rich primitive and composes it with a small fixed library of analytic primitives for grounding, staging, transport, navigation, and release. Rather than expanding the skill library, the harness learns the operating range of these fixed primitives from task-specific execution traces, global success rules, and failure models. By lifting semantic re-grounding, non-contact execution, and VLA re-staging to the planner while reserving the frozen VLA for local contact-rich phases, Harness VLA extends pretrained VLAs beyond their original trajectory distribution without finetuning. Across perturbed tabletop, household kitchen, and clean-to-randomized bimanual manipulation, Harness VLA improves over the strongest relevant baselines by 38.6 and 25.4 percentage points on LIBERO-Pro and RoboCasa365, respectively, and reaches 58.4% on RoboTwin C2R.

Persistent AI agents extend large language models (LLMs) beyond single-turn interaction into long-lived software systems. Unlike traditional chat assistants, unsafe content in these agents can propagate through persistent state, reusable skills, and tool-mediated interactions, creating a substantially larger semantic attack surface. We observe that most security-critical interactions in such agents are transmitted through natural-language token flows, including memory updates, tool arguments, retrieved files, and inter-component communications. This observation enables a new security formulation: unsafe behavior can be intercepted as risky semantic flows before reaching privileged runtime sinks. Based on this insight, we propose TokenWall, a runtime defense framework that acts as a semantic firewall over agent token flows. TokenWall performs boundary-aware semantic auditing over these flows, constructing structured source-sink audit records, applying lightweight local inspection before execution, and selectively escalating ambiguous high-risk cases to stronger arbitration modules. Unlike prior approaches that rely on sparse auditing or remote large-model oversight, TokenWall enables full-coverage pre-execution mediation while reducing remote arbitration and latency. Experiments on CIK-Bench show that TokenWall reduces attack success rate to 12.5% while maintaining a 97.4% benign executable pass rate without human confirmation. TokenWall further introduces only 0.69 seconds of additional latency on benign cases, demonstrating that semantic runtime containment can achieve a practical security-utility trade-off for persistent AI agents.

Amortized neural inverse design typically remains closed-world: component choices are fixed vocabulary tokens, coordinate grids are frozen at training time, and continuous variables are discretized into sequence tokens. Multilayer optical coatings are an industrially important instance, coupling material sequence, layer thickness and wavelength-dependent response. We present IrisFlow, a query-based, open-vocabulary flow-matching framework instantiated in coatings: the target reflectance/transmittance spectrum, wavelength grid, candidate-material optical constants and layer count are supplied at query time. Candidate materials enter as wavelength-aware optical tokens rather than learned identities; material sequences are sampled by discrete flow matching over the query's candidate bank, thicknesses by continuous flow matching without discretization. A single 136M-parameter model designs 2-100-layer stacks. Across a 224-task benchmark it reconstructs in-distribution targets faithfully and retains same-order accuracy on a 15-material held-out bank without retraining; it reconstructs bands up to 1100 nm beyond its training envelope, designs against analytic application specifications and outperforms an autoregressive baseline on that baseline's material library. With optical constants calibrated to our deposition process, IrisFlow designs four color-displaying coolers, fabricated by ion-assisted evaporation: the three chromatic devices reach a CIEDE2000 color error of 3.1-5.2 while retaining 93-95% solar near-infrared reflectance, demonstrating open-vocabulary design carried through to fabricated coatings.

Vietnam's ethnic minority languages are almost absent from the field of Natural Language Processing (NLP), and the challenge goes beyond data scarcity: Cham, Khmer, and Tay-Nung differ sharply in script, Vietnamese contact, and standardization, conditions under which standard multilingual adaptation can learn the wrong signals. We introduce CKTN, the first corpus and benchmark for these languages (44,367 documents, 24M subword tokens), spanning continued pretraining, category classification, and summary-document retrieval. We show that existing multilingual encoders severely fragment these languages, and that common adaptation metrics can mislead: models may lower language-modeling loss or excel at lexical-overlap retrieval while still failing at semantic generalization across documents. We address this with a script-aware adaptation recipe - vocabulary augmentation combined with calibrated replaced-token pretraining - that prevents the discriminator from exploiting trivial script mismatches. The result is an encoder with substantially less fragmentation and the strongest classification performance among evaluated models, exposing the limits of lexical-overlap retrieval as an evaluation signal.

Learning transferable visuomotor imitation policies that generalize across diverse manipulation tasks and adapt rapidly to new tasks from only a handful of demonstrations remains challenging. Most modern policies are trained end-to-end to map observations directly to low-level actions, offering little explicit structure for reusing and recombining behaviors across tasks and making transfer data-inefficient under limited supervision. We propose SkillPlug, a plug-in framework that augments an existing visuomotor policy with a skill-conditioning module and mines a shared, transferable skill library from raw multi-task demonstrations. SkillPlug learns skills via self-supervised objectives that promote compact, reusable, and non-redundant behavior-level primitives, forming a task-shared prior for compositional control. After skill mining, we keep the learned skills fixed and specialize to unseen tasks by fine-tuning only lightweight router and action head, enabling efficient adaptation without full end-to-end retraining. We evaluate SkillPlug on two simulation benchmarks and on a real robot, and observe that the mined transferable skills consistently improve both multi-task performance and few-shot adaptation. Overall, SkillPlug offers a scalable way to mine reusable skills that improve data-efficient generalization in robotic manipulation.

Teleoperation is a key interface for controlling dexterous robotic hands and collecting demonstrations for imitation learning. Its effectiveness largely depends on kinematic retargeting, which maps operator hand motions to feasible and intuitive robot hand motions. Existing methods often require hand-crafted objectives, precise calibration, or global shape matching between human and robot hand spaces, making them sensitive to hand-specific tuning and less reliable across different dexterous hands. We propose AnyDexRT, a calibration-free retargeting method for intuitive dexterous teleoperation across human-like dexterous hands. AnyDexRT combines self-supervised fingertip correspondence learning with few-shot human guidance to anchor the mapping in task-relevant regions, and further refines pinch-related poses using a contact classifier. Experiments on diverse dexterous hands and real-world teleoperation tasks show that AnyDexRT improves retargeting quality, reduces manual tuning, and provides more intuitive and efficient control than prior retargeting methods. Project website: https://chenxi-wang.github.io/projects/anydexrt

Financial markets are noisy, non-stationary, and high-dimensional, making it difficult to discover predictive and robust trading signals. Alpha discovery has evolved from manual factor design to machine learning, evolutionary search, and recent LLM-based frameworks, improving the efficiency of factor generation, search, and evaluation. However, existing methods still mostly automate isolated steps, rather than functioning as end-to-end quant researchers that can absorb external knowledge, close the hypothesis-to-code validation loop, and learn from accumulated discovery feedback. To fill this gap, we introduce XAlpha, a memory-driven AI Quant Researcher for continuous hypothesis-to-code alpha discovery. XAlpha maintains a multi-source research memory system that integrates report-grounded financial knowledge with discovery feedback from prior generations and research cycles. Guided by this memory system, a Macro Brain plans research themes and selects suitable Archetypes; a Micro Brain transforms the planned hypothesis pool into executable factor code and verifies ex-ante tri-alignment among the hypothesis idea, code logic, and financial plausibility; and a Cross Brain consolidates empirical outcomes into generation-level feedback, cycle-level summaries, and archetype-level research cues for future exploration. In this way, XAlpha turns alpha mining from isolated factor generation into a closed-loop research process that continuously reads, hypothesizes, implements, validates, reflects, and evolves. Experiments on CSI300 show that XAlpha achieves stronger overall alpha discovery performance than representative baselines.

Vision--Language--Action (VLA) policies such as $π_{0.5}$ and OpenVLA perform well on many manipulation tasks, but they are often reactive: the next action is predicted from the current observation, instruction, and proprioceptive state. This assumption breaks down in stage-dependent manipulation, where visually similar states may require different actions depending on latent task progress and previous interaction outcomes. We argue that such tasks require not only memory, but dynamics-aware belief updates: the policy should preserve task progress during stable or occluded phases and revise its belief near contact, release, or subgoal transitions. We introduce Temporally Conditioned Memory-Fusion Policies (TFP), a lightweight memory-action framework for VLA backbones. TFP maintains an episode-local task-progress belief with Liquid Time-Constant dynamics and injects the updated belief directly into the flow-matching action decoder through adaptive modulation. This lets temporally accumulated context shape the generated action chunk, rather than serving only as passive history context. With a 3.3B-parameter model, TFP improves the average success rate from \(96.9\%\) to \(98.75\%\) on LIBERO and from \(91.4\%\) to \(93.77\%\) on LIBERO-plus. On the memory-focused MIKASA ShellGameTouch diagnostic, TFP achieves success up to \(75.0\%\). Mechanistic analyses show that write-gain changes near manipulation events are about \(6\times\) larger than far non-event phases, and hidden-state interventions show that the belief causally modulates generated action chunks. These results suggest that compact, event-sensitive memory dynamics can improve VLA policies under occlusion, visual perturbation, and stage-dependent task structure.

While Large Language Models (LLMs) have become essential productivity tools, their integration into workflows without adequate safeguards creates significant risks. This paper proposes an open-source, privacy-focused, user-facing firewall designed to secure both web-based and programmatic LLM interactions. The architecture combines a browser extension and a proxy for total traffic interception across both HTTP(S) and WebSocket communications. At its core, a flexible multi-agent pipeline delivers data leakage prevention through a hybrid approach combining deterministic detectors with LLM-driven semantic analysis, proprietary code leakage prevention, and extensible components designed for future security enhancements such as prompt injection evasion. The framework's layered architecture enables deployment across heterogeneous environments, allowing organizations to balance computational cost, detection depth and latency. Evaluation results demonstrate it achieves F1 scores of up to 94.93% on optimal configurations.

Photonic integrated circuits (PICs) have emerged as a key platform for information processing, including optical communication and computing. As PIC complexity increases, fabrication-induced response deviations accumulate, making post-fabrication trimming critical for unlocking their full potential. Here, we demonstrate that silicon ion implantation enables scalable, ultra-low-loss trimming of cladded PICs by locally forming a high-index silicon-rich region. Structural characterization confirms that the silicon-implanted nanoregion is confined to the cladding without observable damage to the underlying waveguide. The implantation-induced excess loss is below 0.001 dB per π phase shift, while the optical response remains stable over a four-month observation period. Automated trimming is demonstrated on a photonic crossbar array, reducing the average channel output variation from 78.5% to 5.9%. These results establish a practical route towards automated, ultra-low-loss post-fabrication trimming for large-scale cladded PICs.

Best-of-$N$ (BoN) inference improves content consistency in zero-shot text-to-speech by selecting from $N$ candidates with an automatic speech recognition (ASR) verifier. We identify an underexplored evaluation confound: a verifier's apparent quality depends strongly on which ASR family judges it. On LibriSpeech-PC test-clean~\citep{librispeechpc} with F5-TTS~\citep{f5tts}, verifier rankings reverse across Whisper, wav2vec~2.0, and HuBERT evaluators, and same-family verifier-evaluator pairs recover 2-3$\times$ more oracle headroom than cross-family pairs despite near-identical representations (linear CKA $0.978$) -- a pattern consistent with identity- or lineage-level coupling rather than representational overlap. We propose two \textbf{cross-family rank ensembles} (rank-averaging and conjunctive max-rank) that attain the lowest mean WER across three independent evaluators -- $1.61\%$ at $N{=}10$ ($-12\%$ relative to F5-TTS) -- with no measurable degradation under automatic SIM-o/UTMOS metrics; the best single verifier drives WER from $2.06\%$ to $1.72\%$ ($-16.5\%$) under the official F5-TTS evaluator. We recommend cross-evaluator triangulation as default reporting practice.

In this paper, we study the automatic schema generation problem: given a collection of historical ship maintenance and operational reports across multiple form categories, automatically discover compact and informative schemas that capture the essential information requirements of each report type. To address this challenge, we propose ASMR, a modular agentic framework consisting of two specialized agents. A Field Generation Agent extracts semantic concepts from historical narratives and generates candidate schema fields through adaptive multi-granularity clustering, while a Structural Optimizer Agent employs reinforcement learning to identify compact, informative, and non-redundant schema representations. The resulting schemas can guide report authors toward producing more complete, consistent, and actionable reports. Preliminary results demonstrate the promise of the proposed approach and highlight several open research challenges at the intersection of data management, agentic AI, and human-centered AI.

Text-to-SQL is a fundamental task in natural language processing that enables users to interact with structured databases using natural language. While large language models (LLMs) have demonstrated remarkable performance on this task, their substantial computational requirements hinder deployment in resource-constrained settings. In this paper, we introduce SQuaD-SQL (Small-Qualified and Distilled for SQL), a novel approach that empowers small language models (SLMs) to approach the performance of LLMs on the Text-to-SQL task while significantly improving efficiency through knowledge distillation and synthetic data generation. Our method comprises three key components: (1) LLM-based synthetic data generation, where structured knowledge is extracted from LLMs via carefully designed prompting strategies; (2) parameter-efficient fine-tuning, enabling full model training on a single consumer-grade GPU; and (3) domain-adaptive fine-tuning, where domain-specific synthetic data further enhances performance in targeted domains. Experiments on the WikiSQL dataset demonstrate that SQuaD-SQL achieves an execution accuracy of 86.9% on the test set, approaching the performance of LLMs while offering faster inference and lower memory usage. These results suggest that, with proper training strategies, SLMs can serve as practical and efficient alternatives for Text-to-SQL applications in resource-limited environments.

Adjoint optimization is a cornerstone of broadband nanophotonic inverse design, but conventional time-domain implementations face a severe memory bottleneck because they retain forward-field histories at every finite-difference time-domain (FDTD) time step. Here, we show that this full time-step storage is unnecessary for band-limited design objectives. By storing forward fields only at Nyquist-compliant temporal intervals and using the resulting sparse field history during the reverse-time adjoint pass, the proposed method enables on-the-fly gradient accumulation without retaining full forward- or adjoint-field histories. This Nyquist-sampled adjoint FDTD framework preserves the two-simulation scaling of time-domain adjoint optimization while substantially reducing the dominant field-storage cost. Gradient verification confirms that Nyquist-compliant sampling reproduces conventional full-storage adjoint gradients with negligible error, whereas undersampling beyond the Nyquist limit produces aliasing-induced gradient degradation. Across four two-dimensional broadband nanophotonic benchmarks and a fully three-dimensional metalens, the method maintains gradient fidelity and optimized device performance while reducing dominant field-storage memory by up to 107x. These results suggest that the principal memory barrier in broadband time-domain adjoint FDTD is not an intrinsic requirement of gradient evaluation, but a consequence of redundant temporal field storage, opening a practical route to large-scale three-dimensional nanophotonic inverse design.

On the recent EyeBench benchmark, predicting reading comprehension from eye movements exposes a stark gap: text-aware models using pretrained language models reach 56--63% AUROC, while gaze-only models operate at chance. We ask how far a gaze-only model can be pushed by lightweight, language-model-free conditioning. Building on the EyeBench AhnCNN baseline, LEXIC-Base, we propose two mechanisms to inject three precomputed word-level difficulty signals, GPT-2 surprisal, word frequency, and word length, into the per-fixation input: direct concatenation, LEXIC-Concat, and a residual mechanism, LEXIC-Res, where a small head predicts typical-reader gaze response and the encoder is conditioned on the deviation. On the OneStop reading comprehension task, with K=5 seed-ensemble training across ten folds, both mechanisms produce statistically consistent AUROC gains on Unseen Text, +1.8 to +2.2 percentage points, Wilcoxon p <= 0.065. LEXIC-Concat additionally lifts Unseen Reader by +2.9 percentage points, p = 0.010. We trace an architectural boundary in LEXIC-Res on Unseen Reader, +1.8 percentage points, p = 0.19, to the prediction head being calibrated to training readers, transferring imperfectly to out-of-distribution readers.

We present the results of HIPE-OCRepair-2026, an ICDAR competition on LLM-assisted OCR post-correction of historical documents. OCR post-correction remains a long-standing challenge in digital heritage: large-scale collections of digitized documents are affected by legacy OCR errors, while re-digitization at scale remains impractical. Large language models (LLMs) offers a major opportunity to revisit this challenge, yet their effectiveness across languages, document types, and noise conditions - and their tendency to hallucinate - remains insufficiently understood. HIPE-OCRepair-2026 pursues two objectives: (i) to evaluate the capabilities of modern OCR post-correction systems, and (ii) to provide a reproducible evaluation framework anchored in the HIPE-OCRepair-2026 dataset, a harmonized multilingual resource consolidating existing and newly curated historical datasets. Participants were tasked with correcting noisy OCR transcripts from historical newspapers and printed works in English, French, and German (17th-20th century), working at the level of coherent transcription units (paragraphs or articles) without access to source images. The evaluation adopts a retrieval-oriented rather than diplomatic scoring approach, reflecting the practical use case of search and access over digitized collections. Four teams submitted systems ranging from zero-shot prompting to continued pre-training and fine-tuning, offering insights into the merits of different adaptation strategies. Results show that modern LLM-assisted systems can significantly improve OCR quality, but performance varies across datasets, languages, and noise levels. Over-correction on low-noise inputs emerges as a recurring challenge, highlighting the importance of evaluation beyond character error reduction. The dataset, scorer, and evaluation pipeline are publicly released to support future research.

Generative video foundation models exhibit strong compositional priors, yet world-action models (WAMs) and video-action models (VAMs) often lose these priors after finetuning on robotic action data. We refer to this discrepancy as the video-action generalization gap. In this paper, we systematically investigate this gap by evaluating a comprehensive design space of VAMs, demonstrating that standard design choices yield no emergent explanation pattern. To explain this behavior, we introduce the Temporal Ratio (TR), an attention-based measure of how strongly the action head relies on future latent rollouts relative to the anchored current frame. TR has two key properties: first, a model's structural reliance on future-predictive latents, measured via TR, acts as a predictor of its compositional generalization capacity; second, it natively fluctuates based on task phase, shifting attention to future frames during planning and reverting to the present frame for precise manipulation. Finally, based on these findings, we propose an inference-time adaptive guidance method, which exploits this intrinsic feature attention pattern to dynamically amplify compositional video conditioning signals precisely when the policy relies on future rollouts. Evaluated on the LIBERO benchmark and real-world tasks, our approach mitigates the OOD-ID compositional generalization gap. More details: https://umishra.me/temporal-ratio/

Contextual biasing seeks to integrate external knowledge into automatic speech recognition (ASR) systems to accurately recognize domain-specific entities. In this paper, we propose COALA (Contextualized ASR Leveraging Biasing Scoring), a robust framework designed to enhance speech-augmented language models (SLMs) in complex multi-entity scenarios. Considering the inherent context-window limitations of SLMs, identifying relevant target entities from a large-scale biasing list is crucial for effective recognition. To this end, COALA maps SLM latent representations into a specialized discriminative space to quantify the matching intensity between audio segments and candidate entities. Furthermore, we address the training collapse in prior study when handling multi-target utterances-where multiple rare words co-occur. Experimental results on the LibriSpeech benchmark demonstrate that COALA consistently achieves superior contextual biasing performance across various biasing list scales.

Event cameras offer microsecond temporal resolution, low latency, and high dynamic range, making them attractive for robotics. However, labeled event-camera data for a specific robot and scene is scarce and expensive to collect, which slows the development of event-based perception and control. We present EVIS: a physics-grounded event camera plugin for NVIDIA Isaac Sim that generates high-rate, fully labeled event streams directly inside a physics simulator. The plugin implements a faithful log-intensity contrast event model with per-pixel asynchronous reference updates; it migrates from a normal RGB camera with few changes and integrates into any Isaac Sim / Isaac Lab scene, inheriting the simulator's physics and frame-perfect ground truth. It is fully configurable, and offers an interpolation option that renders only sparse keyframes and synthesizes the in-between frames through bidirectional motion-vector warping, making real-time generation on a single GPU possible. Optional sensor noise and motion blur further narrow the gap to real cameras. The generated streams are directly usable by pretrained event networks for downstream tasks. Code repository: https://github.com/spikelab-jhu/isaac-sim-event-camera-plugin

Large language models (LLMs) increasingly act as integrated data-science agents, combining abstract reasoning with advanced tool use. Yet the relevant benchmark landscape largely divides into symbolic causal reasoning benchmarks without realistic data analysis or data analysis benchmarks without a principled causal data-generating structure. Furthermore, existing causal evaluation datasets are often restricted to curated examples from existing sources, with diversity coming from limited templatized variations rather than from systematic generation of novel synthetic causal structures. We introduce CausalDS, a benchmark for evaluating causal reasoning in agentic data-science workflows. Each benchmark instance is a scene consisting of a sampled structural causal model (SCM) with generated observational data and an accompanying synthetic natural-language story grounded in a realistic domain. We optionally ground the composition of the benchmark components in empirical distributions obtained from real-world datasets, thus retaining empirical structure while reducing the "causal parrot" risk through completely synthetic generation. From each scene, we then derive tasks spanning all three of Pearl's rungs, with typical data-science prediction tasks appearing as Rung 1. Most tasks include a data science coding component, where the model typically needs to use several tools to arrive at the final answer due to the frequent presence of imperfect observations, which are generated by an observation model. Additionally, recognizing when a question admits no warranted answer and abstaining is treated as a first-class scored outcome. The benchmark thus jointly evaluates symbolic causal reasoning, data science, uncertainty quantification, abstention, and tool use/coding.

Aspect Sentiment Triplet Extraction (ASTE) requires jointly identifying (aspect, opinion, sentiment) triples from a given review sentence. While large language models (LLMs) achieve strong zero-shot performance on many NLP benchmarks, their effectiveness on ASTE remains limited, as single-pass generation forces the model to determine span boundaries, opinion grouping, and sentiment polarity in a single decoding step. Common remedies, such as few-shot in-context learning and chain-of-thought prompting, offer only marginal improvements and rely heavily on either in-domain demonstrations sampled from labeled training data or carefully engineered reasoning prompts, neither of which is broadly available in zero-shot deployment. Inspired by the classical agent paradigm, we propose MASTE, a multi-agent pipeline for zero-shot Aspect Sentiment Triplet Extraction. MASTE decomposes ASTE into four sequential stages, where specialized agents handle different compositional subtasks with explicit conditioning on prior outputs. This design enables entirely training-free zero-shot ASTE and generalizes across different backbones and datasets. Extensive experiments on four ASTE benchmarks show that MASTE substantially outperforms zero-shot and chain-of-thought LLM baselines under the same backbone, narrowing the gap to fully supervised methods without using any labeled triplets. Code is available at https://github.com/Hankerlove/MASTE.

End-to-end models that map multimodal inputs directly to future trajectories/maneuvers have emerged as an increasingly prominent research paradigm in autonomous driving. This class of models includes both Vision-Language-Action models and trajectory-generative planners. Unlike classic machine learning applications, autonomous vehicles operate in safety-critical and interaction-intensive environments where traditional open-loop imitation of expert demonstrations is not sufficient to ensure reliability. In particular, small execution errors can accumulate over time, while recovery behaviors are scarce in training data. In addition, long-horizon objectives such as safety and driving comfort are not captured by pointwise labels either. These limitations have motivated a shift toward post-training techniques, which further refine driving policies beyond pure imitation. This survey presents a unified view of post-training for autonomous driving by defining its scope and organizing the existing literature into four major families based on the form of supervision they use. For each family, we discuss its capabilities, limitations, and open challenges. We aim to facilitate a systematic understanding of this emerging area and stimulate future research on reliable and efficient post-training for autonomous driving.

Large language models (LLMs) are used worldwide, yet disproportionately reflect Western values, limiting their ability to represent diverse value systems. We introduce PLURAL, a large-scale, value-focused preference dataset grounded in the Integrated Values Survey (IVS), a nationally representative survey spanning 92 countries. Using a two-stage generation pipeline, we transform survey responses into synthetic preference triplets that preserve normative value signals while producing realistic scenarios. We release an initial version of PLURAL containing ~500,000 preference triplets representing people in 20 diverse countries. We evaluate PLURAL in three ways: (i) dataset-level validation showing that it preserves both cross-country value differences and within-country diversity from the original survey; (ii) automated evaluation showing that training on PLURAL improves alignment with target countries' cultural profiles, reducing mean absolute error by up to 27.7% relative to strong baselines; and (iii) blind human evaluation with 176 evaluators in India, Brazil, and Japan, who judge PLURAL-aligned responses as more representative of their national values. Together, these results show that PLURAL contains learnable signal for value steering, offering a scalable resource for pluralistic alignment. Dataset: https://huggingface.co/datasets/agdhruv/plural-alignment

Enterprise large language model (LLM) applications often begin as prototypes whose behavior is carried by prompts and retrieval context. Productization adds requirements for source boundaries, entity routing, answer contracts, and reproducible traces. We present a harness-engineering approach that reconstructs this pattern into a traceable, auditable LLM-agent architecture: deterministic behavior moves into code, manifests, schemas, and validation artifacts around a replaceable composition boundary, while source-backed claims remain the authority for runtime answers. We instantiate it on a public-data slice of five Korean corporate groups (25 listed companies) and evaluate three research questions. (1) The harness preserves its source-grounding, entity-routing, trace, output-hygiene, and recommendation-language contracts across the fixed validation scenarios; a fault-injection control confirms the validators flag deliberately broken contracts. (2) The checks the harness enforces held under model substitution: across three hosted models, they passed on all 270 composition-boundary runs; failures were confined to the model-composed side and were caught and recorded. (3) The code-owned guarantees are load-bearing, not reproducible by prompting alone: holding the model fixed and varying only the enforcement layer, prompt instructions alone let recommendation-language and internal-trace-leakage violations reach the reader, which the harness blocks entirely. A bolt-on external guardrail prevents such violations too but over-refuses, dropping utility to 88/120 where the harness preserves full utility (120/120); in this ablation, only code-owned enforcement preserves both safety and utility. The result is a reusable engineering pattern for turning exploratory prototypes into auditable applications with versioned source, control, and validation artifacts.

Production LLM agents often waste latency and reliability by regenerating code for the same procedural steps on every request. We replace this inference-time coding loop with an agentic tool-making pipeline that compiles repeated SOP steps into validated, versioned tools before deployment. The tool-maker grounds synthesis in the live environment as it collects execution traces, observes backend schemas and values, generates candidate tools, and repairs them against labeled cases. At runtime, the production agent calls these tools directly and falls back to code generation only when needed. We deploy the approach in a Fulfillment Center alarm-triage system, where an agent diagnoses alarms against a 44-node SOP over heterogeneous metric backends. In production, tool calls reduce p50 latency by 42%. On 1,500 historical alarms, they reduce end-to-end error rate by up to 53% by suppressing run-to-run variance in repeated steps. Because tools return compact structured verdicts, they also enable a simpler direct-call architecture, reducing p50 latency by a further 62% in a controlled ablation. Versioned tools also improve auditability and expose specification gaps and upstream data drift. Our results show that self-evolving agents can make industrial LLM systems faster, more reliable, and easier to operate.

We introduce a Bloom-aligned framework for measuring educational control in Large Language Models (LLMs): the ability to preserve a task's instructional intent while shifting its cognitive demand toward specified learning objectives. We apply this framework to programming tasks in computer science education to study the gap between solving tasks and adapting them for learners. Using revised Bloom's Taxonomy as an operational scale of cognitive demand, we evaluate two intervention settings: general difficulty control, where models are asked to make tasks harder or easier, and Bloom's control, where models are asked to target higher or lower Bloom's levels. We evaluate a matched Qwen3-Next model pair, comparing Qwen3-Next-80B-A3B-Instruct with Qwen3-Coder-Next across 2,520 tasks from three benchmarks. The framework reveals a robust directional asymmetry: both models reliably increase cognitive demand, but struggle to lower it. We further characterize these outcomes with semantic-delta clustering and layer-wise Fisher's Discriminant Ratio probing. Within this controlled comparison, the general model shows clearer middle-layer separability for both general difficulty and Bloom-control contrasts, whereas the coder model shows weaker separability for general difficulty and a deeper peak for Bloom-control contrasts. These results show that strong execution performance does not automatically entail Bloom-aligned educational control.

Identifying faithfulness hallucinations in LLM-generated outputs remains challenging due to the scarcity of high-quality annotated data. Recent work relies on advanced LLMs to synthesize training data, including rationales, labels, and hallucinated claims. However, these methods treat the generator as a static component, limiting iterative improvement of the detector. To address this limitation, we introduce Hallucination Self-Play (HSP), a novel framework that enables the detector to bootstrap with an evolved generator. HSP involves two roles initialized from the same base model, a detector that assesses the faithfulness of model outputs, and a generator that produces increasingly hard-to-detect hallucinated responses. Specifically, the detector is first fine-tuned on human-labeled data and then employed as a reward model to train the generator via reinforcement learning from AI feedback (RLAIF). In turn, the evolved generator synthesizes hallucination data to further optimize the detector through rule-based reinforcement learning. Experiments on RAGTruth benchmark and two model families demonstrate that the proposed framework can progressively enhance a small LLM to match or even outperform advanced LLMs without external supervision. Our code is available at https://anonymous.4open.science/r/Hallucination-Self-Play-50B5 .

Large language model (LLM) based multi-agent systems enable complex problem solving through coordinated reasoning and action, but their distributed structure also introduces new challenges in diagnosing system-level failures. When an execution fails, identifying which agent is responsible and at what point the trajectory first becomes irreversibly misdirected is difficult due to long-horizon interactions and tightly coupled agent behaviors. In this paper, we study the problem of failure localization in LLM-based multi-agent systems and present AgentLocate, a framework that attributes failures to both a specific agent and the earliest decisive step. AgentLocate combines an LLM-based judging mechanism with multi-perspective verification by independent evaluators, whose assessments are aggregated using a confidence-aware strategy. The resulting feedback is further used to adapt the judge through lightweight fine-tuning, improving attribution quality. We evaluate AgentLocate on two complementary benchmarks covering diverse tasks, agent configurations, and trajectory lengths. Experimental results show that AgentLocate consistently outperforms existing failure localization methods in identifying both responsible agents and failure steps, while remaining efficient in terms of token usage and running time.

We report the empirical reliability of Gemini models as audio judges that score full-duplex agent conversations directly from the raw stereo waveform, tested across three models in the Gemini family: 2.5 Flash, 3.5 Flash, and 3.1 Pro. Our primary evidence base uses Gemini 2.5 Flash as the ground-truth model, validated against three calibrated human raters on 209 stereo sessions, scored on 8 production dimensions: 152 full-duplex conversations across 13 accent-and-condition strata, together with 57 adversarial defect-injected clips. The evidence for Gemini 2.5 Flash is consistent across three tests. (i) On 5 of 8 dimensions the LALM-human Spearman rho departs from the pairwise human-human rho by at most 0.07, and on 7 of 8 dimensions the two quantities 95 percent bootstrap confidence intervals overlap. (ii) The LALM agrees with the three-rater human mean within 1 point on 60 to 92 percent of sessions on 6 of 8 dimensions. (iii) On 45 of 48 (defect, dimension) cells the LALM is as sensitive as humans or better under Newcombe-Wilson 95 percent confidence intervals, though most of these are underpowered nulls rather than demonstrated parity. Rank-ordering ability transfers across the Gemini family: 3.5 Flash improves simple agreement to 8 of 8 dimensions, while 3.1 Pro rates several dimensions markedly lower than humans despite comparable rank correlation. A model swap should be re-validated on calibration specifically, not assumed from rank-correlation alone. We identify four areas where deployment requires care, and we estimate that human rating alone for our current evaluation cadence costs roughly two orders of magnitude more than the equivalent LALM workload. The data presented here provides a defensible empirical basis for deploying the LALM as a substitute or fourth rater on the dimensions where the evidence supports it.

Reinforcement learning (RL) has achieved remarkable success in enhancing the reasoning capabilities of large language models (LLMs). However, widely used critic-free RL methods rely on uniform credit assignment, broadcasting the same advantage to all tokens regardless of their differences. We identify a critical failure mode of this design, which we refer to as Positive-Credit Contamination: low-probability tail tokens that are contextually erroneous receive identical positive credit to plausible ones within the same trajectory, resulting in the indiscriminate reinforcement of flawed reasoning behavior. To mitigate this issue, we propose Tail-Aware Credit calibratiOn (TACO), a method that calibrates uniform credit assignment to suppress undesirable positive updates. TACO first computes a tail-risk score that incorporates the local generation context to assess each token's risk of falling into the unreliable tail, distinguishing unexpected rarity from uncertainty-driven exploration. TACO then uses this score to tune positive credit for risky tokens without removing their gradients entirely, so that recurring useful rare patterns can accumulate reinforcement while incidental noise is progressively dampened. Experimental results across three LLMs and eight benchmarks show that TACO consistently outperforms GRPO-style baselines. Notably, TACO improves training stability, supporting sustained performance gains in long-horizon RL. The source code is available at: https://github.com/xiuyilou/TACO.

Recent advances in actuation, control and learning have rapidly pushed humanoid robots from a distant vision towards near-term real-world deployment. Healthcare is a particularly pressing domain, in which staffing shortages and increasing care demand are widening the gap between clinical workload and available skilled labour. Although current automation has largely focused on digital and logistical tasks, much hospital work remains embodied, requiring mobility, manipulation and safe interaction in human-designed environments. Humanoid form factors offer unique potential, particularly for assisting with surgical tasks. Traditionally, robotic systems for surgery are purpose-built platforms such as Intuitive Surgical's da Vinci Surgical System, and it remains unclear how close current humanoid systems are to meeting the precision, control and safety requirements of minimally invasive surgery. Here we present a systematic evaluation of contemporary humanoid technology for laparoscopic surgical tasks. We develop a humanoid-based laparoscopic teleoperation framework using general-purpose instruments and assess its abilities through benchtop characterization, dry-laboratory user studies spanning diverse surgical experience levels and in vivo porcine studies. Across these evaluations, we quantify technical feasibility, task performance and clinical readiness relative to established surgical platforms. Together, our study provides an evidence-based assessment of current humanoid abilities and limitations for surgical applications, highlighting both their promise and key technical challenges that must be addressed before clinical deployment.

Soft robotic exogloves can provide hand rehabilitation and assistance. Fitting these gloves often relies on standardized measurements not tailored to the individual, limiting their effectiveness, especially for fine articulation necessary for dexterous manipulation. We present the design, fabrication, modeling, and testing of a personalized pneumatically-actuated soft robotic exoglove. The glove was fit to a user's hand with topological scans and fabricated with silicone mold casting. Finite element analysis (FEA) was performed to evaluate actuator bending and forces from physical human-robot interaction (pHRI) between an actuator and a simplified personalized biomechanical finger model. Pneumatic pressure control experiments were conducted to flex the user's finger with static and dynamic references. Fabrication results show that topological scans enable precise tailoring to hand anatomy. Simulations showed that anatomical personalization enables analysis of pHRI contact forces, and results indicate sufficient joint mobilization with non-ideal compression on the proximal phalanx. Pneumatic testing indicates that pressure control allows accurate and targeted mobility of the metacarpophalangeal (MCP) and proximal interphalangeal (PIP) joints with intrinsic stiffness. Testing of multiple designs showed that relaxing the strain-limiting layer improves actuator-to-finger joint alignment during actuation. This work presents personalization to the human hand in structural conformability, joint topology, modeling of pHRI contact, and time-dependent actuation-deformation profiles. This lays a groundwork for informing exoglove design optimization to enable assistance in dexterous manipulation and neuromuscular rehabilitation of fine motor skills.

Hand spasticity and resulting pain affect 12 million people worldwide, including stroke survivors, arthritis patients, and those with other muscle and nerve deficiencies. Soft robotic exogloves are being introduced to help patients enhance mobility or manage pain; however, there are no current solutions that address both pain and mobility. We present preliminary development of a soft robotic exoglove that both aids in mobility and administers massage-like compression to relax spastic muscles. The glove consists of soft pneumatic actuators that are personalized to an individual's hand topology and kinematics, allowing for optimal conformability and targeted mobility. Novel soft actuators were designed, analyzed, fabricated, assembled into an exoglove, and experimentally tested. Actuators were 3D modeled and analyzed with finite element modeling under pressures of 100 and 200 kPa. Geometries were optimized to minimize stress before fabrication and testing. A dorsal finger actuator was successfully customized to a participant's hand topology, providing full conformal contact and maximal force distribution. A ventral finger actuator was successfully fabricated that can be drastically compressed in size to fit into the tight space of a hyperflexed spastic finger. A palmar actuator was successfully printed with stereolithography, showing potential for 3D-printed soft actuators with more complex geometries. The glove was assembled and successfully worn by a pilot user to validate initial findings in comfort and effectiveness.

Motion and emotion are core parts of intelligent, expressive behavior. In this paper, we introduce fog, a function composition framework for implementing and compose motion functions. We demonstrate how fog can be used to express motion and emotion in Heider-Simmel style animations. This code generation framework can help users generate functions for verbs, adverbs, gestures, and emotions to create an open-ended motion vocabulary. It is complemented by an animation editor that helps users refine motion through direct manipulation and dynamically generated UI. We evaluate our approach with a perceptual evaluation, where we test 452 fog-generated animations to see if people can recognize the semantic meaning of the motion. We find that fog's motion functions can be recognized at 68% accuracy, a 2.68x improvement over a chance baseline. In a mixed-methods user study with professionals and novices, we show that fog in interface form can support users with more rapid iteration, exploration, and control.

As Green Software Engineering matures, energy efficiency has transitioned into a mission-critical non-functional requirement. While software design patterns ensure structural integrity, their inherent abstraction layers impose an implicit "metabolic cost" that often remains obscured during the design phase. This paper empirically investigates the energy dynamics of the Memento design pattern, contrasting a direct, unabstracted baseline against Classic full-snapshot and Differential delta-encoding strategies. Leveraging the RAPL interface for high-fidelity hardware telemetry, we quantify energy dissipation across state volumes scaling from 10 MB to 200 MB. Our empirical results expose a critical architectural trade-off: the Differential strategy minimizes memory traffic, yielding a maximum energy reduction of 65.8% for mid-scale states, but collides with a catastrophic "memory wall" at 200 MB. At this saturation point, algorithmic optimizations are completely neutralized by severe GC thrashing and non-linear power spikes. We synthesize these findings into evidence-based heuristics, providing architects with a robust framework to reconcile structural design quality with sustainable Green IT imperatives.

Preprocessing-based methods for stereotype mitigation, such as pre-/post-training on debiased corpora, are widely used in NLP. While these approaches reduce measurable stereotypes for targeted groups, we find they often induce unintended shifts-side effects, where stereotyping or counter-stereotyping can increase relative to neutral baselines for other demographics, including across unrelated demographic categories. We demonstrate these side effects across two model families (encoder-only and decoder-only), multiple preprocessing strategies (removing stereotypical sentences, removing group mentions, and swapping group references), and both pre- and post-training at different data scales on Wikipedia. Standard benchmarks frequently miss these shifts. Using attention-rollout analysis, we observe that such side effects are not accompanied by large changes in attention flow, complicating mechanistic explanations. We discuss implications for evaluation, provide actionable diagnostics, and argue for side-effect-aware, transparent mitigation practices.