Search arXiv⌕ Search

arXiv subjects

Chen Li

Publications and source records attributed to Chen Li.

At least 19 recordsLinked to original sources
arXiv2026

H-GRPO: Permutation-Invariant Reinforcement Learning for Improving Spatial Grounded Reasoning

Vision-language models (VLMs) have achieved strong performance on visual reasoning tasks, yet a correct final answer does not necessarily imply that the model followed a correct or visually grounded reasoning process. We study how VLMs can be trained to produce explicit intermediate reasoning steps that are both semantically meaningful and grounded in visual evidence. We represent each reasoning step as a structured triplet consisting of a sub-question, sub-answer, and supporting bounding box, enabling the reasoning process to be evaluated along both semantic and spatial dimensions. To optimize this representation, we introduce Hungarian-GRPO (H-GRPO), a reinforcement-learning framework that aligns predicted and reference reasoning steps using permutation-invariant Hungarian matching, avoiding the assumption that valid reasoning steps must follow a fixed ordering. Experiments show that optimizing grounded reasoning improves both intermediate reasoning quality and answer accuracy, with an H-GRPO variant achieving the strongest overall grounding and decomposition performance. Our analyses show that grounding quality and answer correctness are related but distinct, while counterfactual evidence masking confirms that the localized regions contain information important to the model's predictions. These findings motivate evaluating VLMs beyond answer accuracy and demonstrate the value of explicit, spatially grounded reasoning.

cs.CV↗
arXiv2026

WeVisOne-VL: Unified Visual Perception from Objects to Pixels

Existing vision-language models (VLMs) remain unreliable for fine-grained visual perception. In this report, we present WeVisOne-VL (WeChat Vision All-in-One), a powerful VLM that unifies object-level and pixel-level perception through structured next-token prediction. It supports multiple visual tasks including object detection, referring localization, counting, OCR, segmentation, and matting, enabling localization at any user-specific granularity, from points and boxes to masks and alpha mattes. Architecturally, we integrate a lightweight pixel decoder into the VLM via a special token to generate instance-level pixel outputs. We train the model with a three-stage curriculum SFT followed by mixed-task RL. Across over 30 benchmarks, WeVisOne-VL demonstrates leading performance on both object-level and pixel-level tasks, while maintaining robust rejection for object-absent queries. By releasing WeVisOne-VL to the open-source community, we hope to advance VLMs from broad visual understanding toward truly reliable perception systems for real-world deployment.

cs.CV↗
arXiv2026

AdaExploreBench: Benchmarking LLM Agents for Proactive Exploration and Test-time Adaptation in Evolving Environments

Large Language Model (LLM) agents are increasingly deployed in unfamiliar and evolving environments, requiring them to proactively explore, understand, and adapt to environmental dynamics. However, existing benchmarks rarely test whether agents can proactively discover how an environment works and adapt as it evolves. To address this gap, we introduce AdaExploreBench, a benchmark for proactive exploration and test-time adaptation in evolving environments. By instantiating environmental dynamics as hidden rules, AdaExploreBench requires agents to actively acquire evidence and infer the rules. Under implicit rule drift, the latent rule changes without announcement, requiring agents to detect from degraded performance that prior knowledge is stale and re-explore. Experiments on eight leading LLMs show that proactive and adaptive exploration remain challenging: GPT-6 Sol achieves only 61.9% calibrated accuracy, while providing sufficient evidence or removing rule drift improves average performance significantly by close to 10%. These results reveal a critical bottleneck in current LLM agents: they can often reason effectively from informative evidence once it is available, but struggle to actively acquire the right evidence and recognize when previously acquired evidence is no longer valid.

cs.CL↗
arXiv2026

Wan-Alpha++: Animating a Single RGBA Asset into Controllable Transparent Video Effects

RGBA video generation is essential for creating compositable visual assets in filmmaking, game development, and digital content production. However, generating high-quality RGB appearance alongside temporally coherent transparency remains a significant challenge. Existing text-to-RGBA video methods often entangle RGB and alpha distributions and offer limited control over the precise appearance of a target asset. In addition, incorporating image conditioning typically suffers from the emergence-disappearance effects, which fail existing first-frame or first-and-last-frame conditions due to their fully transparent boundary frames. To address these issues, we present Wan-Alpha++, a unified framework that brings controllable RGBA video generation to both text- and image-to-video diffusion backbones through a shared RGBA latent space, extending our conference version Wan-Alpha from the text-conditioned setting to image-conditioned generation with an interior reference frame. Our method first learns shiftable RGB-alpha distributions while preserving the pretrained RGB prior. Specifically, we introduce a transparency-aware bidirectional diffusion objective to produce distinguishable RGBA latent representations, and a transparency-guided mean sampler that shifts the noise distribution to improve transparency generation and enable spatial controllability. Building upon this RGBA generation prior, we introduce temporal phase reparameterization for image-conditioned effect generation. By cyclically transforming a zero-positive-zero alpha pattern into a positive-zero-positive pattern, it relocates visible content to both temporal boundaries, facilitating effective appearance conditioning with limited RGBA data.

cs.CV↗
arXiv2026

Sensitive dependence of the motion of a legged robot on granular media

Legged locomotion on flowing ground ({\em e.g.} granular media) is unlike locomotion on hard ground because feet experience both solid- and fluid-like forces during surface penetration. Recent bio-inspired legged robots display speed relative to body size on hard ground comparable to high performing organisms like cockroaches but suffer significant performance loss on flowing materials like sand. In laboratory experiments we study the performance (speed) of a small (2.3 kg) six-legged robot, SandBot, as it runs on a bed of granular media (1 mm poppy seeds). For an alternating tripod gait on the granular bed, standard gait control parameters achieve speeds at best two orders of magnitude smaller than the 2 body lengths/s ($\approx 60$ cm/s) for motion on hard ground. However, empirical adjustment of these control parameters away from the hard ground settings, restores good performance, yielding top speeds of 30 cm/s. Robot speed depends sensitively on the packing fraction $ϕ$ and the limb frequency $ω$, and a dramatic transition from rotary walking to slow swimming occurs when $ϕ$ becomes small enough and/or $ω$ large enough. We propose a kinematic model of the rotary walking mode based on generic features of penetration and slip of a curved limb in granular media. The model captures the dependence of robot speed on limb frequency and the transition between walking and swimming modes but highlights the need for a deeper understanding of the physics of granular media.

physics.bio-ph↗
arXiv2026

Undulatory Swimming in Sand: Subsurface Locomotion of the Sandfish Lizard

The desert-dwelling sandfish (Scincus scincus) moves within dry sand, a material that displays solid and fluidlike behavior. High-speed x-ray imaging shows that below the surface, the lizard no longer uses limbs for propulsion but generates thrust to overcome drag by propagating an undulatory traveling wave down the body. Although viscous hydrodynamics can predict swimming speed in fluids such as water, an equivalent theory for granular drag is not available. To predict sandfish swimming speed, we developed an empirical model by measuring granular drag force on a small cylinder oriented at different angles relative to the displacement direction and summing these forces over the animal movement profile. The agreement between model and experiment implies that the noninertial swimming occurs in a frictional fluid.

physics.bio-ph↗
arXiv2026

Anomalous Thermal Conductivity Reversal Driven by a Fluctuating Charge Density Wave

For more than a century, the lattice thermal conductivity of crystalline solids has been known to decrease with increasing temperature, as the phonons that carry heat scatter more strongly via Umklapp processes. Here we report a pronounced violation of this rule in the layered charge-density-wave (CDW) metal 2H-TaSe2. Its in-plane thermal conductivity is strongly nonmonotonic: it peaks sharply at the 122 K ordering transition, collapses to a deep minimum near 210 K, and then anomalously reverses, rising with temperature to at least 350 K. Such a high-temperature reversal, in which a crystal conducts heat more efficiently as it is heated, is rare in crystalline solids. Its thermal conductivity minimum lies deep in the disordered phase well above the ordering transition, unlike the weak, transition-centered anomalies of other charge-density-wave systems, pointing to heat conduction governed by fluctuating rather than established order. Combining inelastic X-ray scattering, selected-area electron diffraction, and synchrotron X-ray diffraction, we attribute the reversal to phonons scattering off persistent, fluctuating charge order that survives more than 150 K above the long-range transition, to at least 300 K. A phenomenological phonon-CDW scattering model, tied directly to the measured diffraction linewidth and intensity, quantitatively reproduces the entire nonmonotonic curve. We also reveal that the charge order melts not abruptly but through a fluctuation- and defect-driven, weak first-order process, leaving short-range correlations far above the transition. This work establishes fluctuating charge order as a sensitive control knob for thermal transport in quantum materials and thermal conductivity as a powerful probe of the hidden charge order.

cond-mat.str-el↗
arXiv2026

Blame Errant Words, Not the Canvas: Defect-Aware Spatial Credit Assignment for Visual Text Generation

Recent reinforcement learning approaches have improved visual text generation, yet generated images still suffer from incorrect wording and malformed glyphs. We identify two coupled limitations: recognition-based rewards often overlook visible glyph defects, while whole-image scalar feedback obscures the uneven quality of individual text regions. These limitations motivate our central principle: $\textit{Blame errant words, not the canvas}$, to effectively preserve local quality distinctions. First, we introduce $\mathbf{TextDART}$, a defect-aware region-targeted reward model that localizes individual text regions and independently evaluates their semantic correctness and glyph quality, providing fine-grained, region-precise feedback. Building on TextDART, we propose $\mathbf{SCANNER}$, which turns these regional judgments into spatially differentiated policy feedback. It normalizes rewards for each text target across sampled images and casts the resulting regional advantages onto the corresponding latent tokens. Extensive experiments on existing benchmarks and our newly established $\textbf{PolyText-Bench}$ demonstrate that our approach achieves state-of-the-art performance in text accuracy and glyph fidelity. Project page: https://sujiehu.github.io/textdart-scanner

cs.CV↗
arXiv2026

Delving into the depths of NGC 3783 with XRISM VI. Evidence for a failed wind based on Fe xxv absorption line variability

NGC 3783 exhibits a prominent Fe XXV absorption trough in the time-averaged XRISM/Resolve spectrum. To investigate its short-term variability, we present a time-resolved analysis of the Fe XXV absorption line during the continuous 10-day observation. We divided the resolved absorption profile into five fixed velocity intervals and measured their normalized equivalent widths in 19 time-resolved spectra. Three velocity intervals exhibit significant equivalent-width (EW) variability. Two velocity intervals respond differently to low- and high-flux states with less emission line contamination. The time-dependent photoionisation modelling is unable to reproduce the most significant observed EW variation ($3.5 \, σ$), possibly owing to uncertainties in the assumed initial equilibrium state, which prevent a precise density measurement. Instead, using photoionisation calculations and assuming the high-density limit, we derive approximate constraints on the ionisation parameters of the two relevant components. The low-velocity component (-320 to -680 km/s) shows a delayed inverse response, indicating a high ionisation parameter, while its relatively short lag timescale indicates a relatively high density and a distance upper limit of 0.05 pc. In contrast, the high-velocity component (-680 to -1040 km/s) manifests a positive response with a longer lag to the ionising continuum, implying a relatively low ionisation parameter. This component appears to be less dense, with an inferred distance upper limit of 0.25 pc. The low-velocity component is consistent with a failed wind, as its outflow velocity is below the local escape velocity, whereas the high-velocity component lies near the escape threshold. Together with the lower-ionisation failed wind detected in 2001, this supports a multiphase failed-wind scenario in the unobscured state of NGC~3783 over more than two decades.

astro-ph.HE↗
arXiv2026

PlanPath: Exploration-Guided Orchestration for Patient-Level Pathology Reasoning

A patient's pathology case spans tens of slides from distinct specimens and anatomical sites, each carrying non-redundant, site-specific evidence. Forming a patient-level conclusion therefore requires locating the slides relevant to a given question and integrating complementary findings across them. However, existing pathology agents analyze one slide at a time and describe morphology rather than measure it, leaving evidence acquisition across a patient's specimens and patient-level quantification underexplored. To address this gap, we propose PlanPath, an agentic framework for patient-level pathology visual question answering (VQA). PlanPath turns the question into an exploration plan that specifies the evidence required, routes to the slides likely to carry it, and drives coarse-to-fine acquisition from tissue architecture down to individual cells. Tissue- and cell-level tools supply the counts, proportions, and extents, which PlanPath aggregates across slides into a traceable answer. To enable evaluation, we build two patient-level VQA datasets for renal and gastric cancer, averaging 15 WSIs per patient, with pathologist-verified question-answer pairs derived from diagnostic reports. Experimental results show that PlanPath outperforms the strongest baseline by 9.4% and 12.1% in overall accuracy on the two datasets and further improves existing slide-level agents when plugged into our framework. The dataset and code are available at: https://github.com/JSGe-AI/PlanPath_1.

cs.CV↗
arXiv2026

SyncReward: Human-Aligned Synchronization Reward for Audio-Visual Generation

Recent years have witnessed rapid progress in joint audio--visual generation. However, audio--visual synchronization remains a challenge, as generated speech and sound often temporally misalign with visual events. A key bottleneck is the lack of human-annotated data, reliable evaluation benchmarks, and synchronization-aware evaluation models for generated audio--visual outputs. To address these limitations, we conduct a comprehensive study spanning data, benchmarking, and reward modeling. We first introduce SyncReward-Data-80K, a dataset containing 80K professional synchronization ratings for clips from eight audio--visual generators. A taxonomy-guided construction pipeline covers four sound-event families: Speech, Onset, Instrument, and Ambient. We further construct \textbf{SyncReward-Bench}, a held-out benchmark of 1K clips, each independently rated by multiple professionals. Using this benchmark, we systematically evaluate specialized synchronization metrics and omni-modal models, revealing their substantial misalignment with human judgments. Finally, we develop SyncReward, a reward model that combines semantic--temporal contrastive pretraining on real clips with human supervision on generated clips. As an evaluation metric, SyncReward achieves the strongest agreement with human ratings among the evaluated methods. As a reward, it improves synchronization through inference-time best-of-$N$ selection and reinforcement learning based post-training. We will release the data, benchmark, model, and code at https://github.com/WeChatCV/SyncReward.

cs.CV↗
arXiv2026

Sensing via whole-body interactions with complex terrains: physical insights from robotic models

In nature, animals interact with complex terrains and often generate forces, deformations, and vibrations across the body and environment. Such interaction-induced mechanical information may guide behavior, yet its generation and use are difficult to resolve because movement rapidly and continually reconfigures the interactions that produce it. Sensory neurophysiology and biomechanics have revealed how mechanical information is transformed within sensory, neural, and musculoskeletal subsystems, whereas studies in complex terrains reveal rich whole-body interactions and behavioral outcomes. Connecting these levels remains a challenge. Here, I present a focused perspective on how biologically guided robotic physical models can help identify candidate mechanical information generated through whole-body interactions, test its physical mechanisms and behavioral benefits, and generate biological hypotheses. Three case studies illustrate this approach, multi-legged locomotor transitions to traverse dense clutter, limbless vertical body bending to propel through uneven terrain, and rapid spider leg crouching to help vibration sensing of prey on a web. Across these systems, correspondence between robots and animals supports the biological plausibility of these candidate physical mechanisms and motivates better measurements and models to determine how the relevant mechanical information is sensed and used to modulate behavior.

physics.bio-ph↗
arXiv2026

Joint Projective Spectrum of $D_{\infty n}$

We compute the joint spectrum of $D_{\infty n}$ with respect to the left regular representation, and finds two generators of the De Rham cohomology group of joint resolvent set which is induced by different central linear functionals.

math.FA↗
arXiv2026

Besov spaces and product BMO over polydisc

We prove that the analytic Besov space $B_{p,p}^{1/p}(\C_+^n)$ embeds continuously into Chang--Fefferman product BMO for every finite $n$ and $1\le p<\infty$. Consequently, every small Hankel operator in a finite Schatten class has a product BMO symbol. The proof uses smooth product wavelets and an estimate for dyadic rectangles. We also give a condition on Fejér means that yields a BMO symbol for a bounded Hankel operator on the polydisc.

math.FA↗
arXiv2026

On $p$-Spread Measures

We study $p$-spread probability measures on the Boolean lattice. We show that if a family of sets $A$ is large under the product Bernoulli-$p$ measure, then no $p$-spread measure can be supported on sets that are not covered by the union of two members of $A$, answering the fractional version of Talagrand's discrete convexity problem. Consequently, we establish a coupling theorem between $p$-spread and product Bernoulli-$p$ measures.

math.CO↗
arXiv2026

SoK: Semantic Decision Engines in Network Control Loops

A semantic decision engine such as Jev can return a valid answer and still miss a network deadline, select an infeasible action or leave the service unverified. We systematize 139 paper families by decision interface, execution path and check ownership. Fifty families claim that their engine fits a control loop or time budget, but only four support the claim with matched measurement. Across all 139, four report deadline attainment. The gap concentrates where the decision has no deterministic computation step. Those 72 families make 22 of the claims, none supported, and name a coverage owner in only two. Bounded tests under one event model show that each gap can reverse an admission verdict. A decision that meets a 10 s budget for every isolated request meets it for none once decisions queue ahead of replayed execution times. The same engine passes one coverage check and fails another. We derive a minimum reporting record, design rules and a research agenda for admitting decision engines to control loops.

cs.NI↗
arXiv2026

Vocabulary-Incomplete Unsupervised Vision-Language Model Adaptation

In certain real-world scenarios, obtaining labeled data is excessively expensive, motivating increasing interest in adapting models without annotations. Vision-language models (VLMs) provide pseudo-labels for unlabeled images through image-text similarity, enabling adaptation without manual labeling. The previous unsupervised vision-language model adaption methods assume access to an adaptation vocabulary containing all class names. However, a complete vocabulary is difficult to construct in real-world scenarios, as unlabeled data may contain classes that are overlooked or fall outside the current task scope. Thus, we propose Vocabulary-Incomplete Unsupervised Vision-language model Adaptation (VI-UVA), where only a subset of class names are available during adaptation. The vocabulary incompleteness induces an adaptation bias toward provided classes, which degrades omitted-class recognition. To address this issue, we propose Label Space Reconstruction (LSR), which reconstructs a latent label space from the unlabeled data and incorporates it into prompt tuning, thereby extending adaptation beyond the incomplete vocabulary and mitigating the adaptation bias. Furthermore, we derive an omitted-class semantic discrepancy bound for LSR and identify the key factors affecting the bound. Extensive experiments across eight benchmarks demonstrate that LSR outperforms existing state-of-the-art methods.

cs.CV↗
arXiv2026

PACE: Policy-Native Adaptive Decision Timing for Long-Horizon Reasoning

Long-horizon reasoning requires deciding not only what actions to take, but how many to execute open-loop before replanning. This number, the execution depth, balances replanning cost against compounding execution errors. Most current systems either fix the execution depth as a hand-tuned scalar or adjust it at inference time with heuristic rules decoupled from the policy; we argue both can be suboptimal. In this work, we treat the execution depth as a learnable, history-conditioned variable of the policy itself, and propose PACE, a model-native VLM policy that jointly predicts what to execute and for how long under a hard decision budget. We evaluate PACE on four long-horizon environments spanning fully and partially observable settings and visual and textual modalities: Sliding Puzzle, Sokoban, ALFWorld, and ScienceWorld. Across all settings, PACE improves success rate by 3.1 to 15.6 percentage points, while using fewer decisions on average, achieving empirical success-decision Pareto dominance.

cs.AI↗