Searcharxiv⌕ Search

arXiv subjects

Haonan Wang

Publications and source records attributed to Haonan Wang.

At least 19 recordsLinked to original sources
arXiv2026

Heterogeneous Distributed Zeroth-Order Nonconvex Optimization with Communication Compression

Distributed zeroth-order optimization is increasingly applied in heterogeneous scenarios where agents possess distinct data distributions and objectives. This heterogeneity poses fundamental challenges for convergence analysis, as existing convergence analyses rely on relatively strong assumptions to ensure theoretical guarantees. Specifically, at least one of the following three assumptions is usually required: (i) data homogeneity across agents, (ii) $\mathcal{O}(pn)$ function evaluations per iteration with $p$ denoting the dimension and $n$ the number of agents, or (iii) the Polyak--Łojasiewicz (P--L) or strong convexity condition with a known corresponding constant. To overcome these limitations, we propose a Heterogeneous Distributed Zeroth-Order Compressed (HEDZOC) algorithm, which is based on a two-point zeroth-order gradient estimator and a general class of compressors. Without assuming data homogeneity, we develop the analysis covering three settings: general nonconvex functions, functions satisfying the P--L condition without knowing the P--L constant, and those with a known constant. To the best of our knowledge, the proposed HEDZOC algorithm is the first distributed zeroth-order method that establishes convergence without relying on the above three assumptions. Moreover, it achieves linear speedup convergence rate, which is comparable to state-of-the-art results attainable under data homogeneity and exact communication assumptions. Finally, experiments on heterogeneous adversarial example generation validate the theoretical results.

math.OC↗
arXiv2026

Charge Filling-Dependent Valley Polarization and its Dynamics Driven by Correlated Interlayer Exciton States in WSe2/WS2 Heterobilayer

Moiré heterostructures of transition metal dichalcogenides provide an emerging platform for studying the strongly correlated electronic and excitonic states. The valley degree of freedom and its dynamics in the strongly correlated electronic and excitonic states in the moiré system are crucial for fundamental understanding and applications of valleytronics. Although the strongly correlated states in moiré systems have been extensively investigated, their impact on valley degrees of freedom and its dynamics remains unexplored. Here, we study the charge filling-dependent valley relaxation dynamics of interlayer excitons (IXs) in WSe2/WS2 moiré heterobilayer system. The evolution of moiré IXs dynamics across the Mott transition is elucidated by combining time-resolved optical spectroscopy with Monte Carlo simulation. Our results show that the radiative decay and recombination rate of IXs are enhanced at the integer filling factors n (=-1). Moreover, the valley relaxation rate of IXs at n=$\pm$1 is significantly decreased. The resulting reduction in wavefunction overlap and phonon scattering process suppresses the nonradiative recombination and intervalley scattering of IXs. These anomalous behaviors are attributed to the modification of IXs recombination pathways and enhance excitonic coherence due to the localization of introduced charge carriers at integer filling of moiré site. Our results explore exciton and valley dynamics as sensitive probes of strongly correlated states in moiré systems and highlight possible ways to control excitonic coherence and valleytronic properties in two-dimensional quantum materials.

cond-mat.mes-hall↗
arXiv2026

Unified Communication Compression Beyond Global Error Bounds for Distributed Nonconvex Optimization

In this paper, we propose a unified compression algorithm for distributed nonconvex opitmization with both the locally- and globally-bounded communication compressors, including 1-bit compressors, saturating quantizers, and the globally-bounded compressors with both relative and absolute compression errors, as well as additional arbitrary bounded noise. We provide a rigorous convergence analysis in nonconvex settings and establish linear convergence under the Polyak-Lojasiewicz (P-L) condition. Notably, we establish an $\mathcal{O}(1/\sqrt{T})$ convergence rate for the locally-bounded class in the distributed nonconvex setting, matching that achieved by the centralized algorithms with 1-bit compressors, where $T$ denotes the total number of iterations. Moreover, one initial uncompressed communication round further yields an order-wise improvement to $\mathcal{O}(1/T^{2/3})$. For the P-L setting and the globally-bounded class, we recover state-of-the-art convergence rates.

math.OC↗
arXiv2026

StreamVibe: Proactive Video Agents for Evolving User Intent

Vision-language models are shifting from offline passive understanding toward online proactive interaction. However, prevailing methods largely rely on explicit requests while overlooking implicit user intents (e.g., "where should I click next?" when a user is navigating a website) and fail to continuously update their understanding as user intents evolve. These limitations make them less suitable for intent-aware interaction in real-world scenarios. To address these challenges, we introduce StreamVibe-Bench, a new benchmark that serves as a testbed for user intent understanding and interaction, comprising 4.1K evaluation samples and 85K training samples spanning explicit, implicit, and dynamic intents. Building on this benchmark, we develop StreamVibe-VL, a streaming proactive model that infers implicit user intents from continuous observations and dynamically adapts its interactions as these intents evolve. We further introduce StreamVibe-Harness, which maintains user intents as persistent states with flexible lifecycles through hierarchical memory, enabling it to track the latest user intents and environmental states. Our experiments reveal three key findings: (i) StreamVibe-VL preserves strong general video understanding capabilities while improving response timing and answer quality by 10.7% over existing baselines; (ii) existing proactive models remain limited in understanding user intents and determining when to respond, especially under implicit and dynamic intents; and (iii) StreamVibe-Harness further improves response consistency and adaptation during sustained interactions, resulting in a 12.3% gain in proactive interaction performance.

cs.CV↗
arXiv2026

Wasserstein convergence rates for empirical measures of point processes

In this paper, we establish upper and lower bounds on the convergence rate of empirical measures of point processes under the Wasserstein distance. To this end, we first introduce a new metric on the space of finite counting measures. Unlike many existing metrics in the literature, our proposed metric incorporates the spatial locations of unmatched points. Drawing on both this new metric and existing metrics from the literature, we identify a set of common structural properties and develop a general framework for Wasserstein distances between point-process distributions based on metrics satisfying these properties. We then use this framework to investigate the convergence rates of empirical measures, providing quantitative guarantees for approximating the distribution of the underlying point process, and further establish corresponding concentration inequalities. These theoretical results may provide ingredients for constructing oracle goodness-of-fit tests for a specified point-process distribution. We also apply our theory to the approximation of Campbell measures and discuss how the resulting bounds can help assess Wasserstein-based learning of generative point-process models.

math.ST↗
arXiv2026

Do Coding Agents Understand Why Systems Crash---and Why They Don't?

Coding agents are increasingly capable of fixing real-world bugs, yet they still struggle with large systems codebases such as the Linux kernel. To understand why, we test how agents reason about execution in these codebases through crash reasoning: predicting from source code whether a given input crashes the system. A correct prediction alone, however, does not establish this ability, since it may come from merely spotting suspicious code. In this paper, we propose source interventions: we pair source trees that differ by a single commit, so that the same input crashes only one of them, and test whether agents predict crashes more often on the crashing tree. Experiments with four coding agents on SysRB, our dataset of 934 such pairs from Linux and OSS-Fuzz projects, yield three main findings. First, agents are biased toward predicting no crash: they rarely raise false alarms but miss many real crashes, and even the best agent misses over a quarter of them. Second, locating the code that determines the crash outcome is not enough: agents often misjudge its effect, especially when the crash occurs far from that code. Third, testing only on crashing code is misleading: agents that merely notice suspicious code predict crashes on both source trees, which would look like correct reasoning on the crashing tree alone. Together, these findings indicate that current coding agents cannot yet reliably reason about execution in complex systems codebases, and that paired evaluation is necessary to measure this ability.

cs.SE↗
arXiv2026

Zero2Repo: Can Coding Agents Build Repositories from Scratch?

Coding agents are increasingly asked to build software rather than patch it, yet benchmarks for from-scratch repository construction are mostly limited to a single language and depend on manually curated tasks. We introduce Zero2Repo, a benchmark in which an agent receives a product requirements document, an interface contract, and an empty workspace, and must deliver a complete repository in the project's native ecosystem. Tasks are produced by a language-agnostic authoring pipeline that converts real, version-pinned open-source projects into behavioral specifications, reproducible environments, and hidden acceptance tests. Each task is validated by execution: a reference implementation derived from the upstream project must pass, and adversarial validation must show that the tests reject incorrect implementations. Evaluation runs production coding agents in isolated containers, withholds the acceptance tests until an explicit submission, and assigns a binary reward only when every test passes, with no LLM judge. The pipeline and harness make no language-specific assumptions and apply to mainstream programming ecosystems; the current release contains Python, TypeScript, Go, and C++ tasks. Even on 11 tasks drawn from repositories that frontier models have very likely seen during training, the strongest agent solves only 10, and every failing submission passes 90-99% of the hidden tests; for the two strongest agents, 67-100% of failed tests trace to a single omission or a low-frequency rule stated in the specification rather than to a missing subsystem, so each failure is a concrete target for improvement.

cs.SE↗
arXiv2026

Each Reward to Its Relevant Tokens: Reward-Specific Token-Level Credit Assignment for Multi-Reward RL

Language models are increasingly trained against several verifiable rewards at once, such as the correctness and the format of a tool call, or the individual constraints of a complex instruction. Reward-decoupled normalization (GDPO) normalizes each reward separately so that different reward combinations keep distinct advantages, but it still spreads every reward's advantage evenly over the whole response, although each verifier checks only a few positions. We propose MRTPO, which assigns the advantage of each reward to the tokens relevant to it, using the information its verifier checks, without auxiliary models or token-level annotations. For each reward, the policy replays its own response with and without that reward's verifier information, and the per-token Jensen-Shannon distance between the two next-token distributions measures each token's relevance and reassigns the advantage. On text-to-SQL, multi-constraint instruction following and tool calling, MRTPO converges faster and achieves better performance than the compared methods.

cs.LG↗
arXiv2026

Synergistic Vision-Language Reinforcement Enables Scalable On-Demand Analysis across Diverse Clinical Tasks

Accurate segmentation of tumors and organs-at-risk is essential for oncology but remains time-consuming and expertise-intensive. Existing AI systems often require manual spatial prompts or task-specific retraining, while generic class labels provide limited guidance for heterogeneous disease targets. Here we present SyRe, a text-prompted segmentation foundation model based on Synergistic vision-language Reinforcement, which strengthens bidirectional interaction between visual and linguistic representations. We introduce Color Region Description (CRD) to generate region-level semantic supervision and construct SyReData, comprising 20 million image-mask-description triplets across 9 modalities and 229 segmentation tasks. Training with diversified prompt formats enables single-target, multi-target and open-ended segmentation, together with invalid-prompt rejection. SyRe achieves accurate segmentation across diverse clinical scenarios, with strong gains on disease-related targets. Evaluation on 28 unseen external datasets, including 17 cancer types, 3 non-cancer diseases and multinational in-house cohorts, demonstrates generalization under distribution shift. Beyond spatial overlap, SyRe-generated masks preserve quantitative pathology information. Across five retrospective tumor imaging cohorts spanning CT and MRI, radiomics derived from these masks stratify survival and improve prognostic modeling over clinical baselines. Clinician-in-the-loop refinement further supports case-level correction. These results establish SyRe as a generalizable framework for scalable quantitative oncology analysis and clinician-guided segmentation.

cs.CV↗
arXiv2026

CLASHarness: Benchmarking Multi-Agent Conflict in Computer Use via a Controllable and Verifiable Generation Pipeline

Vision-language models (VLMs) are increasingly deployed in multi-agent systems to automate complex computer-use workflows, yet existing benchmarks mostly assume consistent, cooperative environments. In contrast, real-world collaboration rarely runs on consistent information: peer or sub-agents frequently report outdated states, issue contradictory instructions, or compete over shared permissions and files, forcing agents to check others' outputs and verify ground truth in the environment. Evaluating such conflict resolution remains challenging: manual authoring cannot scale across complex systems, while unconstrained LLM generation produces narrative drift, answer leakage, and unsatisfiable rubrics. We introduce CLASHarness (Conflict-Loaded Agent Scenario Harness), a controllable framework for generating verifiable multi-agent conflicts in computer use, with a benchmark of 600 multi-round tasks. By combining hierarchical scenario planning with an eight-step verified micro-pipeline, CLASHarness pre-commits decisive visual facts before rendering and compiles hybrid code and semantic graders directly from executed reference trajectories, guaranteeing every task is solvable by construction. Evaluating 15 frontier models on CLASHarness reveals three core bottlenecks. First, when sub-agent outputs conflict with the environment, weaker models blindly follow misleading information, whereas stronger models inspect visual evidence to verify ground truth. Second, resolving conflicts is much harder than finding them: while agents can spot inconsistencies, actively fixing system state drops task success significantly. Third, coordination protocols remain a universal bottleneck: even frontier models fail over 58% of closing actions, often forgetting to release shared locks or hand off tasks. CLASHarness provides a scalable, verifiable foundation for building reliable multi-agent systems.

cs.MA↗
arXiv2026

Lost in Adaptation: Layer-Selective Recovery of Temporal Reasoning in Video-Language Models

Multimodal adaptation can weaken temporal reasoning (TR) in video-language models (VLMs) even when the visual evidence needed for an answer is correctly perceived. This weakened competence, however, remains partly recoverable from the model's paired pre-adaptation language checkpoint. We introduce MERIT, a gradient-free framework that selectively restores this competence across the language backbone while protecting temporal perception (TP). It assigns each self-attention layer a VLM-dominant or LLM-dominant interpolation and uses the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to search the recipe space under a perception-constrained objective. Across three VLM families and five video benchmarks, MERIT improves reasoning-aligned performance in all 15 model--benchmark evaluations while preserving TP across all three Video-MME settings. Recipes selected using 232 diagnostic examples transfer to four unseen benchmarks, with relative gains up to 27.8%, and outperform full-backbone, uniform-attention, and alpha-tuned random-layer merging. Interventional masking shows that MERIT-selected layers are disproportionately important for TR, while frame-level attribution reveals increased reliance on temporally distributed, causally relevant evidence. These results show that adaptation-induced reasoning degradation can be repaired post hoc through selective reuse of pre-adaptation competence, without learning new parameters or adding inference overhead.

cs.CV↗
arXiv2026

Better with Experience: Self-Evolving LLM Agents for Evidence-Grounded Health Community Notes

Large Language Model (LLM)-augmented Community Notes offer a scalable path for timely, evidence-grounded correction of health misinformation on social platforms. However, they still reset at every post, leaving useful correction experience from prior cases unused. We introduce EvoNote, an agentic framework that enables health Community Notes generation to self-evolve through an evolving experience memory of prior misinformation correction episodes. Its core is fine-grained credit assignment: EvoNote grounds trajectory-level feedback in health-specific note qualities and distills it into action-level memory for claim analysis, evidence acquisition, and note writing. We evaluate EvoNote on MM-HealthCN, a 1.2K-instance multimodal benchmark of user-flagged health posts with human-written Community Notes and crowd-derived helpfulness labels. Under a human-validated hierarchical utility judge, EvoNote-generated notes are preferred over corresponding human-written notes in 89.6% of cases; on a separate set of Needs More Ratings posts without a crowd helpfulness verdict, EvoNote produces helpful notes for 82.0% of cases. It also reduces time to an available candidate correction from multi-hour post-to-note delays to under 2 minutes on average. Analyses link these quality gains to stronger evidence use and reusable correction strategies, positioning self-evolving note generation as a promising paradigm for scalable health misinformation governance.

cs.CL↗
arXiv2026

Neurons++: A Unified Visual-Stream Framework for Image and Video Reconstruction from fMRI with Neural-to-Visual Interpretability

Decoding visual experience from fMRI is important for understanding human perception and developing brain-computer interfaces. Recent generative decoding methods can reconstruct plausible images and videos, but visual plausibility alone does not ensure that the output is faithfully grounded in brain activity. In this work, we propose Neurons++, a visual-stream fMRI reconstruction framework for both static image and dynamic video reconstruction. Compared with our conference version Neurons, Neurons++ extends the paradigm from video-only reconstruction to visual-stream reconstruction for both images and videos, and replaces parallel task conditioning with serial cortex-inspired decoding. The model first uses image-fMRI pretraining to obtain a stable spatial-semantic initialization, and then performs subject-specific fine-tuning for image and video reconstruction separately. The serial design organizes fMRI-derived representations into explicit spatial, semantic, language-level, and temporal cues. Extensive experiments show that Neurons++ achieves state-of-the-art video reconstruction and competitive image reconstruction across standard benchmarks. Beyond reconstruction fidelity, neural-to-visual analyses indicate that Neurons++ exhibits qualitative directional controllability under neural perturbation, that its fMRI encoder read-out is sensitive to stimulus-relevant cortical features, and that it supports ROI-guided semantic modulation. Code and model weights are available at: https://github.com/xmed-lab/NeuronsPlusPlus.

cs.CV↗
arXiv2026

Convergence of Practical Muon

Muon is emerging as a promising alternative to AdamW for large-scale neural network training, yet theoretical understanding of its practical implementation remains incomplete, as existing analyses often simplify or omit two key components: (i) practical Newton--Schulz iterations with empirically tuned polynomial coefficients $(3.4445,-4.7750,2.0315)$; and (ii) decoupled weight decay for regularization. In this paper, we provide an optimization interpretation and establish convergence for practical Muon, jointly accounting for both components. Specifically, we interpret practical Muon as right-preconditioned optimization of the original loss with a dynamic weighted $\ell_2$ regularizer that vanishes as stationarity is approached, so that the optimization target remains the original objective. We then establish, to our best knowledge, the first convergence guarantee for practical Muon in the stochastic nonconvex setting, with an $\mathcal{O}(T^{-1/4})$ convergence rate in terms of the expected Frobenius norm of the gradient, improving the dimension dependence of the best known AdamW's convergence rate by a factor of $\sqrt{d}$, where $T$ is the iteration horizon and $d$ is the parameter dimension. Experiments further support the theoretical convergence results.

cs.LG↗
arXiv2026

1% of Tokens Can Be Enough: On Gradient Estimation in On-Policy Distillation

Sparse on-policy distillation (OPD) allocates teacher supervision to a small subset of tokens in student-generated trajectories. However, useful teacher guidance can yield a noisy update when its gradient is estimated from a sampled next token. We study this estimation problem at a fixed prefix in information geometry and propose an information-efficiency ratio (IER) based on a signal-to-noise decomposition. IER characterizes relative gradient estimation error under an optimal scalar baseline. A candidate-set approximation enables token selection based on IER and its combination with existing usefulness scores, while retaining the sampled reverse-KL training objective. On mathematical and medical reasoning tasks, adding IER improves existing selectors in multiple settings, with sparse configurations matching or exceeding full OPD without token selection at small token budgets of 0.1\%--1\%. These results support accounting for both usefulness and gradient-estimation reliability when allocating sparse supervision. Our code is available at https://github.com/BruceSheng1202/IER-OPD.

cs.LG↗
arXiv2026

Prioritized Rollouts for Efficient World Model-based Vision-Language-Action Policy Optimization

Vision-Language-Action (VLA) models have emerged as a powerful paradigm for embodied intelligence, but fine-tuning them with reinforcement learning (RL) remains constrained by the cost of real-world robot interaction. Model-based reinforcement learning (MBRL) reduces this cost by using a learned world model to generate rollouts for policy optimization. However, it becomes computationally expensive as VLA policies and world models scale. Existing methods typically treat states equally, overlooking substantial differences in their utility for policy improvement. In this paper, we show that policy uncertainty helps identify states with greater potential for policy improvement. The policy exhibits high uncertainty at only a small subset of states, often during decision-sensitive stages where small action differences can alter task outcomes, suggesting that policy improvements at these states could be particularly valuable. Building on these findings, we introduce U-GROW, a lightweight, plug-and-play sampling layer that directs more model rollouts to these informative states. By modifying only the branched-start distribution, U-GROW can be integrated into existing MBRL pipelines without changing the policy optimization objective. Experiments in both simulated and real-world manipulation tasks demonstrate the efficiency and effectiveness of U-GROW, supporting the use of policy uncertainty to guide experience generation.

cs.LG↗
arXiv2026

KHiM-Mamba: Injecting Pathology Knowledge into Mamba via Hidden-State Modulation for Whole Slide Image Analysis

Whole slide image analysis is commonly formulated as multiple instance learning (MIL), where instance features are contextually updated and aggregated into a slide representation, a process we term slide encoding dynamics. Recently, selective state-space models (SSM) have emerged as promising MIL architectures due to their long-sequence modeling capability and linear complexity. However, existing SSM-based MIL methods rely solely on visual features during MIL. Meanwhile, in large-scale WSIs, where sparse diagnostically decisive regions are surrounded by abundant irrelevant information, such purely vision-driven selective dynamics can misallocate state updates and readouts, causing the evolving SSM state to accumulate task-irrelevant evidence and dilute critical diagnostic cues over long scan trajectories. In this work, we propose the Knowledge-Aware Hidden-State Modulation architecture (KHiM-Mamba), which innovatively regulates Mamba's core selective state-space mechanism with explicit knowledge priors, steering slide encoding dynamics toward diagnostically meaningful evidence accumulation. Specifically, we redesign the original SSM layer to perform knowledge modulation operations during the evolution of hidden states, thereby guiding what visual evidence is accumulated and retrieved from the hidden state at each encoding step. Furthermore, we additionally introduce a local-adaptive vocabulary retrieval module that uses large language models to assign each patch fine-grained, tissue-specific semantic descriptions, enabling precise modulation across diverse tasks. Experiments on 11 public benchmarks across 4 tasks show that KHiM-Mamba consistently achieves state-of-the-art performance.

eess.IV↗
arXiv2026

Efficient Discrete Position Design for Movable Antenna Systems: Low Complexity and Robustness

Building on advances in reconfigurable antenna techniques, movable antennas (MAs) can dynamically reshape antenna arrays and introduce additional spatial degrees of freedom (DoFs), thereby further improving communication performance. Despite these benefits, existing MA design algorithms often entail prohibitively high computational complexity from discrete positioning selection, which prevents practical implementations of MAs. In this paper, we investigate efficient solutions for the mutual information (MI) maximization problem of a multi-user multiple-input multiple-output (MU-MIMO) uplink communication system aided by discrete MAs. To this end, we first formulate the discrete MA positioning problem with the assumption of perfect channel state information (CSI). Then, we prove that the design problem falls into the category of monotone submodular maximization subject to a 2-system constraint. Accordingly, we propose a low-complexity distance-constrained submodular position search algorithm, which is theoretically shown to achieve at least 1/3 of the optimum. Furthermore, we extend our approach to scenarios with imperfect CSI, and show that the proposed submodular optimization-based design remains robust against channel estimation errors. Numerical results demonstrate that the proposed scheme can achieve at least 90% of the optimal solution's MI gain under both perfect and imperfect CSI assumptions. Remarkably, the algorithm achieves orders-of-magnitude complexity reduction (e.g., 34.4x faster than the branch-and-bound approach) while maintaining significant MI gains.

cs.IT↗