GraspMolmo predicts semantically appropriate, stable grasps conditioned on a natural language instruction and a single RGB-D frame. For instance, given "pour me some tea," GraspMolmo selects a grasp on a teapot handle rather than its body. Unlike prior TOG methods, which are limited by small datasets, simplistic language, and unrealistically simple scenes, GraspMolmo learns from PRISM, a novel large-scale synthetic dataset of 379k samples featuring complex environments and diverse, realistic task descriptions. We fine-tune the Molmo vision-language model on this data, enabling GraspMolmo to generalize to novel open-vocabulary instructions and objects. In challenging real-world evaluations, GraspMolmo achieves state-of-the-art results, with a 70% prediction success on complex tasks, compared to the 35% achieved by the next best alternative. GraspMolmo also successfully demonstrates the ability to predict semantically correct bimanual grasps zero-shot. We release our synthetic dataset, code, model, and benchmarks to accelerate research in task-semantic robotic manipulation, which, along with videos, are available at this URL.

Node and link churn in multi-party, cross-region clusters over wide-area networks (WANs) often disrupts distributed training. However, checkpoint-based recovery and cloud-centric autoscaling react slowly and assume centralized control, which is misaligned with the self-governed setup where institutions can freely join and leave. This paper proposes Chaos, a multi-party distributed training system with self-healing and autoscaling, enabling robust and elastic training under churn. It speeds up autoscaling via multi-neighbor state replication and model sharding. We formalize the sharding and assignment as a MINLP that captures WAN heterogeneity, and reduce it to a tractable MILP by analyzing its monotonicity on a divisibility chain. By establishing an equivalence, we derive a greedy algorithm that follows optimality rules and yields the optimal solution in polynomial time. Chaos uses a cluster monitor to track resource and topology changes, and handles scaling events through peer negotiation protocols, enabling fully self-governed autoscaling among institutions. Experiments show that Chaos has substantially lower scale-out delay than Pollux, Elan, and Autoscaling, and handles scale-in, connect-link, and disconnect-link events within 20ms. It also delivers the lowest idle time, showing superior resource use and scalability as the cluster grows.

Potential failures of physics-informed machine learning in traffic flow modeling: theoretical and experimental analysis Yuan-Zheng Lei a, Yaobang Gong a, Dianwei Chen a, Yao Cheng b, Xianfeng Terry Yang* a a University of Maryland, College Park, MD 20742, United States b Florida Atlantic University, Boca Raton, FL 33431, United StatesAbstract This study investigates why physics-informed machine learning (PIML) may fail when it comes to macroscopic traffic flow modeling. We define failure as the case where a PIML model underperforms both its purely data-driven and purely physics-based counterparts by a given threshold. Our analysis shows that physics residuals themselves do not inherently hinder the optimization of the loss function, which is a main reason responsible for the failure of the PIML model in other fields. Instead, successful parameter updates require both machine-learning and physics gradients to form acute angles with the true gradient. Our experiment shows that this condition may be hard to achieve for PIML under a general low-resolution loop dataset. In particular, when the traffic data resolution is low, a neural network cannot accurately approximate density and speed, causing the constructed physics residuals, already affected by discrete sampling and temporal averaging, to lose their ability to reflect the actual PDE dynamics. This degradation can directly lead to PIML failure. From a theoretical standpoint, we show that although the exact solutions of the LWR and ARZ models are weak solutions, for piecewise C k initial data and under mild conditions, the solutions remain C k on the complement of the shock set over finite time, with only finitely many shock waves, where C k refers to k times continuously differentiable. Since the shock set has Lebesgue measure zero, the probability of a detector measurement or auxiliary collocation point lying exactly on a discontinuity is essentially zero; asymptotically, every auxiliary point admits a sufficiently small smooth neighborhood where the physics residual is well-defined and valid. Consequently, the well-known limitation that MLPs cannot exactly represent non-smooth functions does not materially affect our setting, as the residual evaluation almost always occurs in smooth regions. We also investigate the error lower bounds of the MSE of physics residuals for PIML models under high-resolution data. We prove that higher-order models like ARZ possess strictly larger consistency error lower bounds than lower-order models like LWR under mild conditions.

Consumer Health Queries (CHQs) in Bengali (Bangla), a low-resource language, often contain extraneous details, complicating efficient medical responses. This study investigates the zero-shot performance of nine advanced large language models (LLMs): GPT-3.5-Turbo, GPT-4, Claude-3.5-Sonnet, Llama3-70b-Instruct, Mixtral-8x22b-Instruct, Gemini-1.5-Pro, Qwen2-72b-Instruct, Gemma-2-27b, and Athene-70B, in summarizing Bangla CHQs. Using the BanglaCHQ-Summ dataset comprising 2,350 annotated query-summary pairs, we benchmarked these LLMs using ROUGE metrics against Bangla T5, a fine-tuned state-of-the-art model. Mixtral-8x22b-Instruct emerged as the top performing model in ROUGE-1 and ROUGE-L, while Bangla T5 excelled in ROUGE-2. The results demonstrate that zero-shot LLMs can rival fine-tuned models, achieving high-quality summaries even without task-specific training. This work underscores the potential of LLMs in addressing challenges in low-resource languages, providing scalable solutions for healthcare query summarization.

Physics-informed deep learning (PIDL) represents a rapidly advancing framework that integrates known governing physical laws, typically formulated as PDEs, into the training process of deep neural networks. In contrast to conventional data-driven models that rely solely on observational data, PIDL incorporates physical constraints to guide learning, thereby enhancing generalization, reducing data dependence, and improving interpretability. This synthesis of physics and deep learning has demonstrated broad applicability in solving forward and inverse problems across scientific and engineering domains, particularly in scenarios involving limited, noisy, or deceptive data. Key methodologies under the PIDL umbrella include physics-informed neural networks (PINNs) [1, 2, 3, 4], which embed PDE constraints via automatic differentiation; sparse identification of nonlinear dynamics (SINDy) [5, 6], which infers governing equations by promoting sparsity in learned representations; and physics-informed neural operators [7, 8, 9, 10, 11], which approximate solution operators across function spaces to model families of PDEs. These approaches are particularly well-suited for high-dimensional problems, where traditional numerical solvers suffer from the curse of dimensionality. High-dimensional PDEs are integral to various scientific and engineering domains, including quantum mechanics, financial mathematics, and optimal control. Their solutions provide crucial insights into complex, multi-scale phenomena that cannot be accurately captured using lower-dimensional approximations. However, solving these equations efficiently remains a significant challenge due to the curse of dimensionality, the exponential growth in computational complexity and data requirements as the number of dimensions increases.

Abstract-- Vision-Language Models (VLMs) have demonstrated significant potential for end-to-end autonomous driving. However, the field still lacks a practical platform that enables dynamic model updates, rapid validation, fair comparison, and intuitive performance assessment. T o that end, we introduce LightEMMA, a Lightweight End-to-End Multimodal Model for Autonomous driving. LightEMMA provides a unified, VLM-based autonomous driving framework without ad hoc customizations, enabling easy integration with evolving state-of-the-art commercial and open-source models. We construct twelve autonomous driving agents using various VLMs and evaluate their performance on the challenging nuScenes prediction task, comprehensively assessing computational metrics and providing critical insights. Illustrative examples show that, although VLMs exhibit strong scenario interpretation capabilities, their practical performance in autonomous driving tasks remains a concern. Additionally, increased model complexity and extended reasoning do not necessarily lead to better performance, emphasizing the need for further improvements and task-specific designs. Autonomous vehicles (A Vs) have seen tremendous advancements over the years, improving safety, comfort, and reliability. Traditional approaches rely on modular designs, rule-based systems, and predefined heuristics [1], [2].

Alignment has quickly become a default ingredient in LLM development, with techniques such as reinforcement learning from human feedback making models act safely, follow instructions, and perform ever-better on complex tasks. While these techniques are certainly useful, we propose that they should not be universally applied and demonstrate a range of tasks on which base language models consistently outperform their popular aligned forms. Particularly, we study tasks that require unpredictable outputs, such as random number generation, mixed strategy games (rock-paper-scissors and hide-and-seek), and creative writing. In each case, aligned models tend towards narrow behaviors that result in distinct disadvantages, for instance, preferring to generate "7" over other uniformly random numbers, becoming almost fully predictable in some game states, or prioritizing pleasant writing over creative originality. Across models tested, better performance on common benchmarks tends to correlate with worse performance on our tasks, suggesting an effective trade-off in the required capabilities.

Multilingual Large Language Models (LLMs) have demonstrated significant effectiveness across various languages, particularly in high-resource languages such as English. However, their performance in terms of factual accuracy across other low-resource languages, especially Indic languages, remains an area of investigation. In this study, we assess the factual accuracy of LLMs - GPT-4o, Gemma-2-9B, Gemma-2-2B, and Llama-3.1-8B - by comparing their performance in English and Indic languages using the IndicQuest dataset, which contains question-answer pairs in English and 19 Indic languages. By asking the same questions in English and their respective Indic translations, we analyze whether the models are more reliable for regional context questions in Indic languages or when operating in English. Our findings reveal that LLMs often perform better in English, even for questions rooted in Indic contexts. Notably, we observe a higher tendency for hallucination in responses generated in low-resource Indic languages, highlighting challenges in the multilingual understanding capabilities of current LLMs.

Large Language Model (LLM) agents have demonstrated impressive capabilities in social deduction games (SDGs) like Werewolf, where strategic reasoning and social deception are essential. However, current approaches remain limited to textual information, ignoring crucial multimodal cues such as facial expressions and tone of voice that humans naturally use to communicate. Moreover, existing SDG agents primarily focus on inferring other players' identities without modeling how others perceive themselves or fellow players. To address these limitations, we use One Night Ultimate Werewolf (ONUW) as a testbed and present MultiMind, the first framework integrating multimodal information into SDG agents. MultiMind processes facial expressions and vocal tones alongside verbal content, while employing a Theory of Mind (ToM) model to represent each player's suspicion levels toward others. By combining this ToM model with Monte Carlo Tree Search (MCTS), our agent identifies communication strategies that minimize suspicion directed at itself. Through comprehensive evaluation in both agent-versus-agent simulations and studies with human players, we demonstrate MultiMind's superior performance in gameplay. Our work presents a significant advancement toward LLM agents capable of human-like social reasoning across multimodal domains.

PEER-REVIEWED AND ACCEPTED IN IEEE- ISTAS 2025 The rapid evolution of Generative AI (GenAI) has introduced unprecedented opportunities while presenting complex challenges around ethics, accountability, and societal impact. This paper draws on a literature review, established governance frameworks, and industry roundtable discussions to identify core principles for integrating responsible GenAI governance into diverse organizational structures. Our objective is to provide actionable recommendations for a balanced, risk-based governance approach that enables both innovation and oversight. Findings emphasize the need for adaptable risk assessment tools, continuous monitoring practices, and cross-sector collaboration to establish trustworthy GenAI. These insights provide a structured foundation and Responsible GenAI Guide (ResAI) for organizations to align GenAI initiatives with ethical, legal, and operational best practices.