arXiv subjects

Yuheng Zhang

Publications and source records attributed to Yuheng Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Labimus: A Simulation and Benchmark for Humanoid Dexterous Manipulation in Chemical Laboratory

Laboratory automation has made remarkable progress through robotic platforms and AI-driven scientific reasoning. However, many laboratory operations (e.g., solid--solid transfer) remain inherently dynamic and require real-time adaptation to different materials and experimental conditions. Such precision-critical manipulations are difficult to standardize, motivating the use of humanoid robots with dexterous hands. Despite this opportunity, no existing benchmark evaluates humanoid manipulation in precision-critical laboratory environments. We present Labimus, to our knowledge, the first benchmark for humanoid dexterous manipulation in organic chemistry laboratories. Labimus reconstructs over 30 functionally faithful assets from real organic chemistry workstations through real-to-sim modeling, collectively covering the core operations of routine organic chemistry experiments. The benchmark integrates articulated laboratory instruments, particle-based powder physics, and closed-loop instrument readouts, enabling a complete manipulation-to-measurement pipeline. It further defines six atomic operations and a seven-step solid-weighing workflow derived from real laboratory standard operating procedures. We introduce a precision-aware evaluation protocol designed to jointly measure task completion, experimental precision, and long-horizon execution. We benchmark three representative policies under procedural layouts and environmental perturbations. Results reveal a precision gap: policies that successfully complete laboratory tasks can still fail to satisfy the quantitative tolerances required by experimental protocols. Our benchmark exposes a fundamental disconnect between task completion and experimental validity, providing a new testbed for developing reliable humanoid robots for scientific laboratories.

cs.RO↗

arXiv2026

TuringViT: Making SOTA Vision Transformers Accessible to All

Modern VLMs and VLA systems commonly adopt off-the-shelf ViTs such as SigLIP2 as visual encoders, but diverse downstream requirements in latency, temporal modeling, and VLM integration often call for customized SOTA-level ViTs. Training such encoders remains beyond the reach of much of the community, as it requires massive image-text data, while standard softmax attention makes high-resolution or dynamic-resolution pretraining prohibitively costly and often forces low-resolution pretraining followed by post-hoc adaptation. TuringViT addresses these challenges with three key designs: Turing Linear Attention (TLA) for efficient sequence modeling, VISTA-Curation to construct supervision-rich image-video training data, and native dynamic-resolution pretraining that supports flexible inputs from the start and transfers seamlessly to downstream VLMs. As a result, TuringViT outperforms leading open-source ViT baselines with only 10% of the data, achieves stronger downstream VLM performance, and delivers substantially better latency scaling on high-resolution inputs. Our scaling-law analysis further shows that TuringViT continues to improve predictably with curated data scale, far from saturation. Its fast adaptation, hardware-friendly design, and efficient deployment have made it a unified visual foundation across XPeng's AI systems. More broadly, TuringViT provides a reproducible pipeline that dramatically lowers the cost for the community to train, customize, and deploy SOTA-level ViTs, moving toward making such Vision Transformers accessible to all.

cs.CV↗

arXiv2026

ULTIMATE-deblending I. A 50-band Ultraviolet to Mid-infrared Photometric Catalog Combining Space- and Ground-based Data in the JWST/PRIMER survey

Our understanding of the early Universe has long been limited by biased galaxy samples selected through various color criteria. With deep JWST infrared imaging, mass-complete galaxy samples can now be studied up to $z \sim 8$ for the first time. However, recent work has revealed systematic uncertainties in measuring physical properties of galaxies based solely on JWST/NIRCam and HST photometry, due to their limited wavelength coverage. This highlights the need for supplementary data, particularly in the rest-frame UV and near-infrared. Here we present the ULTIMATE-deblending project, which will eventually deliver self-consistent UV to radio photometry for galaxies detected in deep JWST surveys, including both NIRCam and MIRI data. In this first paper, we release a 50-band photometric catalog spanning CFHT/U to JWST/MIRI F1800W, covering a total of 627.1 arcmin$^2$ across two JWST/PRIMER fields. We detail the reduction of the JWST imaging data, the photometric procedures, and the spectral-energy-distribution-fitting methodology used to derive the galaxy properties. Compared with photometry including only HST and JWST bands, the inclusion of deblended low-resolution photometry from ground-based telescopes improves the accuracy of photometric redshifts by $\sim$40%, while reducing the outlier fraction by $\sim$60%. This galaxy sample can serve as a key reference for statistical studies of galaxy formation and evolution in the early Universe. The UV-to-MIR catalogs and JWST mosaics from the ULTIMATE-deblending project have been made publicly available.

astro-ph.GA↗

arXiv2026

Spider-Webb: enhanced star formation in low-mass galaxies within the Spiderweb protocluster revealed by JWST Pa$β$ narrow-band imaging

Understanding the role of the environment in galaxy evolution is key to revealing the physical processes that regulate galaxy growth. We study the star formation activity of \pab\ emitters (PBEs) in the Spiderweb protocluster at $z=2.16$ using \textit{James Webb Space Telescope}/NIRCam narrow-band imaging. To investigate the environmental dependence of star formation, we derive star formation rates (SFRs) from the \pab\ emission line and compare SFRs in the Spiderweb protocluster with those in the field. Our main finding is that low-mass PBEs ($M_\star < 10^9\,M_\odot$) in the Spiderweb protocluster exhibit an enhancement in star formation compared to their field counterparts. This excess persists even without applying dust-attenuation corrections, indicating that enhanced star formation in the protocluster is robust regardless of whether a dust correction is applied. In contrast, intermediate- and high-mass PBEs ($M_\star > 10^9\,M_\odot$) show no significant deviation from the field, revealing a strong mass dependence in the environmental effects on star formation. No clear spatial concentration toward the cluster core of starbursting low-mass galaxies within the protocluster is seen, suggesting that their enhancement is not restricted to the cluster core. We suggest that starbursts in low-mass galaxies are facilitated by environmental processes such as galaxy mergers/interactions, and/or efficient gas supply. While the enhancement at the low-mass end is consistent with trends reported for other protoclusters at similar redshifts, the behaviour of star formation at intermediate masses ($10^{9} < M_\star/M_\odot < 10^{10}$) is not uniform across protoclusters. Our \pab-based results in the Spiderweb protocluster indicate that star-formation enhancement at cosmic noon depends on both mass and the dynamical state of the protocluster.

astro-ph.GA↗

arXiv2026

Pessimism-Free Offline Learning in General-Sum Games via KL Regularization

Offline multi-agent reinforcement learning in general-sum settings is challenged by the distribution shift between logged datasets and target equilibrium policies. While standard methods rely on manual pessimistic penalties, we demonstrate that KL regularization suffices to stabilize learning and achieve equilibrium recovery. We propose General-sum Anchored Nash Equilibrium (GANE), which recovers regularized Nash equilibria at an accelerated statistical rate of $\widetilde{O}(1/n)$. For computational tractability, we develop General-sum Anchored Mirror Descent (GAMD), an iterative algorithm converging to a Coarse Correlated Equilibrium at the standard rate of $\widetilde{O}(1/\sqrt{n}+1/T)$. These results establish KL regularization as a standalone mechanism for pessimism-free offline learning that achieves equivalent or accelerated rates in multi-player general-sum games.

cs.LG↗

arXiv2026

Fast Rates in $α$-Potential Games via Regularized Mirror Descent

An $α$-potential game is a multi-player non-cooperative interaction in which a global potential function approximates individual player rewards up to a structural bias $α$. While identifying a Nash Equilibrium (NE) in generic general-sum games is known to be computationally intractable, the potential game structure enables tractable NE identification. In this paper, we study the offline learning of NE in $α$-potential games using KL regularization. To analyze this process, we propose a novel Reference-Anchored offline data coverage framework--a verifiable condition that anchors data requirements to a known reference policy rather than an unknown optimum. Building on this, we propose Offline Potential Mirror Descent (OPMD), a decentralized algorithm that achieves an accelerated $\widetilde{\mathcal{O}}(1/n)$ statistical rate, surpassing the standard $\widetilde{\mathcal{O}}(1/\sqrt{n})$ rate typical of offline multi-agent learning. This work characterizes the first fast-rate offline learning approach for $α$-potential games.

cs.GT↗

arXiv2026

A Python-Based Peeling Framework for Radio Interferometry: Application to uGMRT 650MHz Imaging

Modern radio interferometric arrays offer high sensitivity, wide fields of view, and broad frequency coverage, but also pose significant data calibration challenges. Standard direction-independent calibration is insufficient to correct direction-dependent effects, such as ionospheric phase distortions and primary beam variations, which produce strong artifacts around bright sources and limit achievable image dynamic range. Built on standard CASA tasks, we present a Python-based direction-dependent calibration and peeling framework, demonstrated using radio continuum imaging data from the upgraded Giant Metrewave Radio Telescope (uGMRT). The framework efficiently subtracts bright-source models and suppresses their associated direction-dependent artifacts, producing significantly flattened backgrounds and improving image fidelity and faint-source detectability. We further introduce an optimized ``model-restoration'' strategy that mitigates direction-dependent artifacts while preserving the flux densities and morphologies of bright sources that are themselves of scientific interest. For fields containing multiple bright sources, sequential application of the framework systematically reduces background noise, thereby increasing sensitivity and faint-source detectability. The framework is Python-based, CASA-compatible, and can be readily applied to other mid- and low-frequency interferometric arrays. The code is publicly released with this paper.

astro-ph.IM↗

arXiv2026

Offline Two-Player Zero-Sum Markov Games with KL Regularization

We study the problem of learning Nash equilibria in offline two-player zero-sum Markov games. While existing approaches often rely on explicit pessimism to address distribution shift, we show that KL regularization alone suffices to stabilize learning and guarantee convergence. We first introduce Regularized Offline Sequential Equilibrium (ROSE), a theoretical framework that achieves a fast $\widetilde{\mathcal{O}}(1/n)$ convergence rate under \textit{unilateral concentrability}, improving over the standard $\widetilde{\mathcal{O}}(1/\sqrt{n})$ rates in unregularized settings. We then propose Sequential Offline Self-play Mirror Descent (SOS-MD), a practical model-free algorithm based on least-squares value estimation and iterative self-play updates. We prove that the last iterate of SOS-MD attains the same $\widetilde{\mathcal{O}}(1/n)$ statistical rate up to a vanishing optimization error of order $\widetilde{\mathcal{O}}(1/\sqrt{T})$ in the number of self-play iterations $T$.

cs.LG↗

arXiv2026

Beyond Pessimism: Offline Learning in KL-regularized Games

We study offline learning in KL-regularized two-player zero-sum games, where policies are optimized with respect to a fixed reference policy through KL regularization. Prior work relies on pessimistic value estimation to handle distribution shift, yielding only $\widetilde{\mathcal{O}}(1/\sqrt n)$ statistical rates. We develop a new pessimism-free algorithm and analytical framework for KL-regularized games, built on the smoothness of KL-regularized best responses and a stability property of the Nash equilibrium induced by skew symmetry. This yields, to our knowledge, the first pessimism-free offline learning guarantee for KL-regularized games, with a fast $\widetilde{\mathcal{O}}(1/n)$ sample complexity bound. We further propose an efficient self-play policy optimization algorithm that replaces exact equilibrium computation with iterative KL-regularized policy updates, and prove that its last iterate preserves the same pessimism-free statistical guarantee up to a controlled optimization error.

cs.GT↗

arXiv2026

Rethinking Importance Sampling in LLM Policy Optimization: A Cumulative Token Perspective

Reinforcement learning, including reinforcement learning with verifiable rewards (RLVR), has emerged as a powerful approach for LLM post-training. Central to these approaches is the design of the importance sampling (IS) ratio used in off-policy policy-gradient estimation. Existing methods face a fundamental bias-variance dilemma: token-level IS ratios, as adopted by PPO (Schulman et al., 2017) and GRPO (Shao et al., 2024), introduce bias by ignoring prefix state distribution mismatch; full sequence ratios provide exact trajectory-level correction but suffer from high variance due to the multiplicative accumulation of per-token ratios, while GSPO (Zheng et al., 2025) improves numerical stability via length normalization at the cost of deviating from the exact full-sequence IS correction. In this work, we identify the cumulative token IS ratio, the product of per-token ratios up to position $t$, as a theoretically principled solution to this dilemma. We prove that, under the token-level policy-gradient formulation, this ratio provides an unbiased prefix correction for each token-level gradient term and has strictly lower variance than the full sequence ratio. Building on this insight, we propose CTPO (Cumulative Token Policy Optimization), which combines the cumulative token IS ratio with position-adaptive clipping that scales log-space clip bounds according to the natural $\sqrt{t}$ growth of the cumulative log-ratio. This yields more consistent regularization across token positions. We implement and evaluate CTPO in the tool-integrated reasoning setting on several challenging mathematical reasoning benchmarks, achieving the best average performance across both model scales compared with strong GRPO and GSPO baselines. Code will be available at https://github.com/horizon-llm/CTPO.

cs.LG↗

arXiv2026

Instructing LLMs to Negotiate using Reinforcement Learning with Verifiable Rewards

The recent advancement of Large Language Models (LLMs) has established their potential as autonomous interactive agents. However, they often struggle in strategic games of incomplete information, such as bilateral price negotiation. In this paper, we investigate if Reinforcement Learning from Verifiable Rewards (RLVR) can effectively teach LLMs to negotiate. Specifically, we explore the strategic behaviors that emerge during the learning process. We introduce a framework that trains a mid-sized buyer agent against a regulated LLM seller across a wide distribution of real-world products. By grounding reward signals directly in the maximization of economic surplus and strict adherence to private budget constraints, we reveal a novel four-phase strategic evolution. The agent progresses from naive bargaining to using aggressive starting prices, moves through a phase of deadlock, and ultimately develops sophisticated persuasive skills. Our results demonstrate that this verifiable training allows a 30B agent to significantly outperform frontier models over ten times its size in extracting surplus. Furthermore, the trained agent generalizes robustly to stronger counterparties unseen during training and remains effective even when facing hostile, adversarial seller personas.

cs.AI↗

arXiv2026

Beyond Semantic Manipulation: Token-Space Attacks on Reward Models

Reward models (RMs) are widely used as optimization targets in reinforcement learning from human feedback (RLHF), yet they remain vulnerable to reward hacking. Existing attacks mainly operate within the semantic space, constructing human-readable adversarial outputs that exploit RM biases. In this work, we introduce a fundamentally different paradigm: Token Mapping Perturbation Attack (TOMPA), a framework that performs adversarial optimization directly in token space. By bypassing the standard decode-re-tokenize interface between the policy and the reward model, TOMPA enables the attack policy to optimize over raw token sequences rather than coherent natural language. Using only black-box scalar feedback, TOMPA automatically discovers non-linguistic token patterns that elicit extremely high rewards across multiple state-of-the-art RMs. Specifically, when targeting Skywork-Reward-V2-Llama-3.1-8B, TOMPA nearly doubles the reward of GPT-5 reference answers and outperforms them on 98.0% of prompts. Despite these high scores, the generated outputs degenerate into nonsensical text, revealing that RMs can be systematically exploited beyond the semantic regime and exposing a critical vulnerability in current RLHF pipelines.

cs.LG↗

arXiv2026

ProOOD: Prototype-Guided Out-of-Distribution 3D Occupancy Prediction

3D semantic occupancy prediction is central to autonomous driving, yet current methods are vulnerable to long-tailed class bias and out-of-distribution (OOD) inputs, often overconfidently assigning anomalies to rare classes. We present ProOOD, a lightweight, plug-and-play method that couples prototype-guided refinement with training-free OOD scoring. ProOOD comprises (i) prototype-guided semantic imputation that fills occluded regions with class-consistent features, (ii) prototype-guided tail mining that strengthens rare-class representations to curb OOD absorption, and (iii) EchoOOD, which fuses local logit coherence with local and global prototype matching to produce reliable voxel-level OOD scores. Extensive experiments on five datasets demonstrate that ProOOD achieves state-of-the-art performance on both in-distribution 3D occupancy prediction and OOD detection. On SemanticKITTI, it surpasses baselines by +3.57% mIoU overall and +24.80% tail-class mIoU; on VAA-KITTI, it improves AuPRCr by +19.34 points, with consistent gains across benchmarks. These improvements yield more calibrated occupancy estimates and more reliable OOD detection in safety-critical urban driving. The source code is publicly available at https://github.com/7uHeng/ProOOD.

cs.CV↗

arXiv2026

RoboMIND 2.0: A Multimodal, Bimanual Mobile Manipulation Dataset for Generalizable Embodied Intelligence

While data-driven imitation learning has revolutionized robotic manipulation, current approaches remain constrained by the scarcity of large-scale, diverse real-world demonstrations. Consequently, the ability of existing models to generalize across long-horizon bimanual tasks and mobile manipulation in unstructured environments remains limited. To bridge this gap, we present RoboMIND 2.0, a comprehensive real-world dataset comprising over 310K dual-arm manipulation trajectories collected across six distinct robot embodiments and 739 complex tasks. Crucially, to support research in contact-rich and spatially extended tasks, the dataset incorporates 12K tactile-enhanced episodes and 20K mobile manipulation trajectories. Complementing this physical data, we construct high-fidelity digital twins of our real-world environments, releasing an additional 20K-trajectory simulated dataset to facilitate robust sim-to-real transfer. To fully exploit the potential of RoboMIND 2.0, we propose MIND-2 system, a hierarchical dual-system frame-work optimized via offline reinforcement learning. MIND-2 integrates a high-level semantic planner (MIND-2-VLM) to decompose abstract natural language instructions into grounded subgoals, coupled with a low-level Vision-Language-Action executor (MIND-2-VLA), which generates precise, proprioception-aware motor actions.

cs.RO↗

arXiv2026

ArtVIP: Articulated Digital Assets of Visual Realism, Modular Interaction, and Physical Fidelity for Robot Learning

Robot learning increasingly relies on simulation to advance complex ability such as dexterous manipulations and precise interactions, necessitating high-quality digital assets to bridge the sim-to-real gap. However, existing open-source articulated-object datasets for simulation are limited by insufficient visual realism and low physical fidelity, which hinder their utility for training models mastering robotic tasks in real world. To address these challenges, we introduce ArtVIP, a comprehensive open-source dataset comprising high-quality digital-twin articulated objects, accompanied by indoor-scene assets. Crafted by professional 3D modelers adhering to unified standards, ArtVIP ensures visual realism through precise geometric meshes and high-resolution textures, while physical fidelity is achieved via fine-tuned dynamic parameters. Meanwhile, the dataset pioneers embedded modular interaction behaviors within assets and pixel-level affordance annotations. Feature-map visualization and optical motion capture are employed to quantitatively demonstrate ArtVIP's visual and physical fidelity, with its applicability validated across imitation learning and reinforcement learning experiments. Provided in USD format with detailed production guidelines, ArtVIP is fully open-source, benefiting the research community and advancing robot learning research. Our project is at https://x-humanoid-artvip.github.io/ .

cs.RO↗

arXiv2026

Interaction-Grounded Learning for Contextual Markov Decision Processes with Personalized Feedback

In this paper, we study Interaction-Grounded Learning (IGL) [Xie et al., 2021], a paradigm designed for realistic scenarios where the learner receives indirect feedback generated by an unknown mechanism, rather than explicit numerical rewards. While prior work on IGL provides efficient algorithms with provable guarantees, those results are confined to single-step settings, restricting their applicability to modern sequential decision-making systems such as multi-turn Large Language Model (LLM) deployments. To bridge this gap, we propose a computationally efficient algorithm that achieves a sublinear regret guarantee for contextual episodic Markov Decision Processes (MDPs) with personalized feedback. Technically, we extend the reward-estimator construction of Zhang et al. [2024a] from the single-step to the multi-step setting, addressing the unique challenges of decoding latent rewards under MDPs. Building on this estimator, we design an Inverse-Gap-Weighting (IGW) algorithm for policy optimization. Finally, we demonstrate the effectiveness of our method in learning personalized objectives from multi-turn interactions through experiments on both a synthetic episodic MDP and a real-world user booking dataset.

cs.LG↗

arXiv2025

WorldMove, a global open data for human mobility

High-quality human mobility data is crucial for applications such as urban planning, transportation management, and public health, yet its collection is often hindered by privacy concerns and data scarcity-particularly in less-developed regions. To address this challenge, we introduce WorldMove, a large-scale synthetic mobility dataset covering over 1,600 cities across 179 countries and 6 continents. Our method leverages publicly available multi-source data, including gridded population distribution, point-of-interest (POI) maps, and commuting origin-destination (OD) flows-to generate realistic city-scale mobility trajectories using a diffusion-based generative model. The generation process involves defining city boundaries, collecting multi-source input features, and simulating individual-level movements that reflect plausible daily mobility behavior. Comprehensive validation demonstrates that the generated data closely aligns with real-world observations, both in terms of fine-grained individual mobility behavior and city-scale population flows. Alongside the pre-generated datasets, we release the trained model and a complete open-source pipeline, enabling researchers and practitioners to generate custom synthetic mobility data for any city worldwide. This work not only fills critical data gaps, but also lays a global foundation for scalable, privacy-preserving, and inclusive mobility research-empowering data-scarce regions and enabling universal access to human mobility insights.

cs.SI↗

arXiv2025

Understanding World or Predicting Future? A Comprehensive Survey of World Models

The concept of world models has garnered significant attention due to advancements in multimodal large language models such as GPT-4 and video generation models such as Sora, which are central to the pursuit of artificial general intelligence. This survey offers a comprehensive review of the literature on world models. Generally, world models are regarded as tools for either understanding the present state of the world or predicting its future dynamics. This review presents a systematic categorization of world models, emphasizing two primary functions: (1) constructing internal representations to understand the mechanisms of the world, and (2) predicting future states to simulate and guide decision-making. Initially, we examine the current progress in these two categories. We then explore the application of world models in key domains, including generative games, autonomous driving, robotics, and social simulacra, with a focus on how each domain utilizes these aspects. Finally, we outline key challenges and provide insights into potential future research directions. We summarize the representative papers along with their code repositories in https://github.com/tsinghua-fib-lab/World-Model.

cs.CL↗