Searcharxiv⌕ Search

arXiv subjects

Xiaohan Wang

Publications and source records attributed to Xiaohan Wang.

At least 37 records · Page 2Linked to original sources
arXiv2026

A quiescent galaxy in a gas-rich cosmic web node at z~3

Recent JWST observations have unveiled a large number of quiescent galaxies at $z\gtrsim3$, bringing potential challenges to current galaxy formation models. Since star formation is expected to be fed by external gas accretion, the knowledge about the circumgalactic media (CGM) of these galaxies is essential to understanding how they quench. In this work, we present the discovery of a massive and passive galaxy ($M_\star\simeq10^{11}\,M_\odot$) within the MQN01 structure at z~3.25, containing one of the largest overdensities of galaxies and active galactic nuclei (AGN) found so far at $z\gtrsim3$. The passive galaxy has a star-formation rate of $4^{+6}_{-2}~M_\odot$/yr, placing it more than 1 dex below the star-forming main sequence, and has no detectable molecular gas ($M_\mathrm{H2}<7\times10^{9}\,M_\odot$). Surprisingly, it is located at the center of a large cool gas reservoir, as traced by bright Ly$α$ and H$α$ emission. By taking advantage of deep multi-wavelength information unique to this field, including deep Chandra X-ray data, we argue that the inefficient gas accretion from the CGM onto this galaxy over the last few hundreds of Myr, as suggested by the observations, could be caused by an AGN jet of a nearby star-forming galaxy located at a projected distance of 48 kpc. In particular, we argue that the jet feedback may have maintained a high level of CGM turbulence around the passive galaxy and thus caused a reduced gas accretion over the required time-scales. In addition, the elevated ionizing field provided by the AGN overdensity, including the nearby AGN, can illuminate the passive galaxy's cool CGM and make it visible through fluorescent emission. Our study demonstrates that the star formation rates of high-redshift galaxies could be substantially reduced and maintained at a low level even within gas-rich and overdense environments in particular situations.

astro-ph.GA↗
arXiv2026

Policy Improvement Reinforcement Learning

Reinforcement learning has become a central post-training paradigm for improving LLM and agent capabilities. Yet existing RL post-training methods share a common blind spot: they construct local learning signals from sampled trajectories, rewards, or feedback-conditioned targets, then update the policy without explicitly verifying whether the resulting policy outperforms its predecessor. Optimizing these local signals does not necessarily produce a better policy, while finite sampling, generation stochasticity and feedback noise can further widen this gap. We argue that the missing ingredient is policy improvement feedback: the ability to measure progress across policy iterations. We introduce Policy Improvement Reinforcement Learning (PIRL), which formulates inter-iteration performance gain as an explicit objective structurally aligned with final task performance. Building on PIRL, we propose Policy Improvement Policy Optimization (PIPO), a plug-in closed-loop framework that verifies the previous update against a sliding-window historical performance anchor. PIPO uses this improvement feedback to modulate the local learning signal of the base policy optimization algorithm, reinforcing updates associated with measured progress and suppressing those associated with performance drops. We provide theoretical evidence that PIPO locally aligns policy updates with the PIRL improvement objective. Experiments on mathematical reasoning, code, tool-use, and self-distillation settings show that PIPO yields consistent gains across PPO, group-relative, and self-distillation policy optimization families.

cs.LG↗
arXiv2026

AutoMem: Automated Learning of Memory as a Cognitive Skill

Memory expertise is a learned skill: knowing what to encode, when to retrieve, and how to organize knowledge--a capacity known in cognitive science as metamemory. We bring this perspective to LLMs by treating memory management as a trainable skill. We promote file-system operations to first-class memory actions alongside task actions, letting the model itself decide how to manage its memory. This memory skill improves along two axes: the structure that supports it (prompts, file schemas, action vocabulary), and the proficiency of the model exercising it. Both axes resist manual optimization: episodes in long-horizon tasks run for thousands of steps, and a single memory mistake can hide long before it surfaces, making human review of full trajectories impractical. We introduce AutoMem, a framework that automates both axes. In the first loop, a strong LLM reviews complete agent trajectories and iteratively revises the memory structure that shapes how the agent interacts with its memory files. In the second loop, the agent's own good memory decisions are identified from many episodes and used as training signal to sharpen the model's memory proficiency directly. Across three procedurally generated long-horizon games (Crafter, MiniHack, and NetHack), optimizing memory alone--without modifying the model's task-action behavior--improved the base agent's performance ~2x-4x, bringing a 32B open-weight model competitive with frontier systems such as Claude Opus 4.5 and Gemini 3.1 Pro Thinking. Our results show that memory management is an independently learnable skill, and a high-leverage objective yielding large gains on long-horizon tasks.

cs.AI↗
arXiv2026

Learning from Failure: Inference-Time Self-Improvement for Computer-Use Agents

Computer-use agents, which leverage multimodal large language models (MLLMs) to operate computers and complete tasks, have attracted significant attention for their utility and versatility. A major challenge in developing these agents is collecting large-scale, high-quality trajectories. The standard approach generates synthetic data through a self-improving loop: an agent is placed in a verifiable environment and iteratively fine-tuned on its successful trajectories. Despite its effectiveness, this paradigm exploits only successful trajectories and discards the failed ones, even though failures carry rich information about a model's weaknesses. In this work, we explore a complementary failure-driven self-improvement loop, a data-centric paradigm that turns failed trajectories into agent improvements. Specifically, we employ an LLM to diagnose failure modes, propose inference-time solutions, and generate code patches -- lightly verified by humans -- that upgrade the agent. We validate this approach with the state-of-the-art OpenCUA-72B model on the OSWorld benchmark, improving the success rate from 42.3% to 48.9%, a gain of 6.6 percentage points, without any additional training cost and with only modest inference overhead. Our results demonstrate that failure-driven self-improvement is a viable complement to success-based pipelines, enabling more efficient agent improvement.

cs.CV↗
arXiv2026

A massive barred spiral galaxy at z = 5.102 discovered by JWST

We report M1149-BSG-z5, a barred spiral galaxy at $z = 5.102$, identified in the parallel field of MACS J1149+2223 with JWST and HST. M1149-BSG-z5 is the highest redshift barred galaxy candidate to date. Both isophote ellipse fitting and structural modeling support a stellar bar of length $a_\mathrm{bar} \approx 4.5$ kpc, and extended spiral arms peaking at $r \approx 5.5$ kpc. M1149-BSG-z5 is a massive main sequence star-forming galaxy, with a stellar mass of $10^{10.45}\rm M_\odot$ and a star-formation rate of $144\,\rm M_\odot/yr$. A concentrated bulge is embedded in an extended disk with a global Sérsic index $n = 2.37$. With an effective radius of $R_{e} = 2.61\rm \ kpc$, M1149-BSG-z5 is larger than typical galaxies at $z \sim 5$ and comparable to barred galaxies at $2 < z < 4$. M1149-BSG-z5 also hosts a broad-line AGN, with a relatively low black-hole-to-stellar mass ratio of $\rm M_{\rm BH}/M_\ast\sim10^{-3}$. Its metal-enriched emission-line properties indicate that it is already chemically evolved. These properties imply M1149-BSG-z5 as an early-assembled and structurally evolved galaxy. We also find that M1149-BSG-z5 resides in an overdense region with a nearby companion galaxy, suggesting an interaction-driven bar formation mechanism. Its concentrated light, early assembly and main-sequence star formation also suggest baryon-dominated, gas-rich conditions, where gravitational instability can further accelerate the bar formation.

astro-ph.GA↗
arXiv2026

TAPO: Tool-Aware Policy Optimization via Credit Transfer for Multimodal Search Agents

We identify and formally characterize credit misassignment as a systematic failure mode of GRPO in tool-augmented multimodal search agents: its uniform broadcast of trajectory-level advantages to all tokens causes valuable tool-use steps in failing trajectories to be penalized no differently from valueless ones. We further empirically quantify the scale of this phenomenon. Over half of failing trajectories and failing tool-use actions exhibit correctable credit misassignment, demonstrating that the wasted training signal is both substantial and structurally exploitable. Building on this insight, we propose Tool-Aware Policy Optimization (TAPO), which exploits the parameter-determinism property of information-acquisition tools: similar call parameters define equivalent information-acquisition actions and should therefore share comparable action credit. TAPO constructs counterfactual witnesses within the current training batch and compensates misassigned negative credit via confidence-gated conservative advantage correction. It requires no additional annotation, models, or sampling, and introduces negligible computational overhead. Across multiple multimodal search benchmarks, TAPO delivers consistent, plug-and-play improvements over strong baselines for three mainstream RL algorithms (GRPO, GSPO, and SAPO). Our code and models will be publicly released upon acceptance.

cs.AI↗
arXiv2026

Are Full Rollouts Necessary for On-Policy Distillation?

On-policy distillation (OPD) provides dense teacher feedback along student-generated rollouts rather than fixed teacher traces and has emerged as a promising post-training paradigm. However, standard OPD typically generates full rollouts during training, which is computationally expensive and may expose the student to unreliable teacher feedback at late rollout positions, especially during early training. We identify the rollout horizon as a key bottleneck in OPD that substantially impacts training efficiency. Unlike Reinforcement Learning with Verifiable Rewards (RLVR), OPD does not require a final answer reward to provide learning signals. Therefore, full rollouts may not always be necessary for OPD. Motivated by this insight, we propose two simple horizon-control strategies: Progressive OPD (POPD), which gradually expands the rollout horizon during training, and Truncated OPD (TOPD), which permanently performs distillation on reliable truncated rollouts. Experiments on mathematical reasoning show that POPD improves the training efficiency of OPD by up to 3$\times$, while TOPD matches OPD performance using only 10\% of the rollout horizon, leading to substantial wall-clock and memory reductions. These results demonstrate that controlling the rollout horizon offers a simple and practical path to more efficient OPD.

cs.CL↗
arXiv2026

LocalSearchBench: Benchmarking Agentic Search in Real-World Local Life Services

Recent advances in large reasoning models LRMs have enabled agentic search systems to perform complex multi-step reasoning across multiple sources. However, most studies focus on general information retrieval and rarely explores vertical domains with unique challenges. In this work, we focus on local life services and introduce LocalSearchBench, which encompass diverse and complex business scenarios. Real-world queries in this domain are often ambiguous and require multi-hop reasoning across merchants and products, remaining challenging and not fully addressed. As the first comprehensive benchmark for agentic search in local life services, LocalSearchBench comprises a database of over 1.3M merchant entries across 6 service categories and 9 major cities, and 900 multi-hop QA tasks from real user queries that require multi-step reasoning. We also developed LocalPlayground, a unified environment integrating multiple tools for LRMs interaction. Experiments show that even state-of-the-art LRMs struggle on LocalSearchBench: the best model (DeepSeek-V3.2) achieves only 35.60% correctness, and most models have issues with completeness (average 60.32%) and faithfulness (average 30.72%). This highlights the need for specialized benchmarks and domain-specific agent training in local life services. Code, Benchmark, and Leaderboard are available at https://localsearchbench.github.io/.

cs.AI↗
arXiv2026

ZipRL: Adaptive Multi-Turn Context Compression with Hindsight Response Replay

Adaptive context compression is vital for scaling Large Language Models (LLMs) to complex, multi-turn agent tasks. However, rule-based compression methods may discard task-critical nuances, while Reinforcement Learning (RL) approaches usually struggle to balance information retention and token efficiency under the sparse rewards inherent to long-horizon workflows. To bridge this gap, we propose ZipRL, a novel adaptive compression framework tailored for Reinforcement Learning from Verifiable Rewards (RLVR). ZipRL features a multi-granularity compression mechanism for active, non-uniform information reduction, coupled with Hindsight Response Replay (HRR), a technique designed to densify training signals during RLVR optimization. Theoretically, we prove ZipRL's superior task-relevant utility over uniform methods. Concretely, ZipRL utilizes coarse-to-fine prompts for macro-compression and incorporates HRR into GRPO via generalized advantage reshaping. Multiple models of varying versions and parameter scales validate the effectiveness of our approach. Benchmarks on five agent tasks show ZipRL outperforms state-of-the-art approaches by 27.9% and 34.7% across Qwen3-4B and Qwen3-8B models, while maintaining exceptional token efficiency and robustness under extreme 256-turn extrapolation stress tests.

cs.AI↗
arXiv2026

Joint Training of Multi-Token Prediction in Reinforcement Learning via Optimal Coefficient Calibration

Reinforcement Learning from Verifiable Rewards (RLVR) has emerged as the standard paradigm for improving reasoning capability of large language models, while Multi-Token Prediction (MTP) has been a widely adopted module in pretraining. Combining them is a natural approach, yet current RL practices detach MTP gradients because joint training degrades the performance. We revisit this failure from an optimization perspective. We show that the per-step effect of MTP on the RL objective can be decomposed into two terms: a first-order correlation and a second-order perturbation penalty. This decomposition unifies three MTP training regimes: Detach, Cross-Entropy loss, and Policy loss, and explains why each succeeds or fails. Further analysis of policy loss reveals that, although it aligns with intuition, performance still degrades: the correlation term decays while the quadratic penalty persists. Guided by the analysis, we propose Optimal Coefficient Calibration (OCC), an adaptive scheme that tracks the optimal coefficient online via a log-probability proxy at negligible cost. Across six competition-level mathematical reasoning benchmarks, OCC consistently matches or exceeds the detach baseline, delivering improved joint MTP-RL training performance.

cs.LG↗
arXiv2026

LiveK12Bench: Have Large Multimodal Models Truly Conquered High School-level Examinations?

Advanced Large Multimodal Models (LMMs) have demonstrated impressive performance in K-12 reasoning tasks, exhibiting great promise as intelligent tutors. Realizing this potential requires models to navigate real-world examinations effectively, yet most existing benchmarks fail to capture the complexity of authentic testing environments. Specifically, most datasets are static, prone to data contamination, and are often confined to restricted modalities, disciplines, and evaluation criteria. To address these issues, we introduce LiveK12Bench, a dynamic, holistic, multi-disciplinary benchmark designed to evaluate the reasoning abilities of LMMs in realistic examination scenarios. LiveK12Bench comprises 2K+ verified questions spanning Mathematics, Physics, Chemistry, and Biology, sourced from the latest real-world exam papers and designed to grow over time. Our framework features several core innovations: 1) featuring an automated pipeline that continuously ingests and parses the latest examination papers to mitigate data leakage; and 2) proposing a novel `Mock Exam' evaluation scheme, which assesses the ability to complete end-to-end exams autonomously with accurate and efficient reasoning paths. Extensive experiments on 12 LMMs reveal that advanced models suffer substantial performance degradation under exam-realistic constraints: GPT-5's score drops from 79 to 53 (out of 100) when process rigor and efficiency are jointly evaluated. Our findings expose critical vulnerabilities, such as sensitivity to complex visual layouts, highlighting the gap between idealized reasoning capabilities and true educational readiness. Both code and dataset are publicly available.

cs.AI↗
arXiv2026

On the Hidden Costs of Counterfactual Knowledge Training in LLM Unlearning

Counterfactual tuning (CFT) has emerged as a promising paradigm for Large Language Model (LLM) unlearning by training models to generate alternative fictitious knowledge in place of undesired content. However, in this work, we find that this paradigm still underperforms other paradigms in some aspects, and identify two previously overlooked pitfalls underlying this gap: (1) knowledge conflict, where mutual inconsistencies within counterfactual corpora induce conflicting gradients that disrupt parameter optimization, and (2) hallucination spillover, where fitting false targets instills a persistent fabrication bias, inflating hallucination rates on unrelated domains. To systematically diagnose these issues, we introduce RWKU+, an extended benchmark equipped with novel trade-off metrics and gradient-level diagnostic tools. Our work further discusses the limitations and overhead of the paradigm, aiming to provide insights and actionable guidance for more rigorous LLM unlearning research.

cs.CL↗
arXiv2026

Contextual Rollout Bandits for Reinforcement Learning with Verifiable Rewards

Reinforcement Learning with Verifiable Rewards (RLVR) is an effective paradigm for improving the reasoning capabilities of large language models. However, existing RLVR methods utilize rollouts in an indiscriminate and short-horizon manner: responses of heterogeneous quality within each prompt are treated uniformly, and historical rollouts are discarded after a single use. This leads to noisy supervision, poor sample efficiency, and suboptimal policy updates. We address these issues by formulating rollout scheduling in RLVR as a contextual bandit problem and proposing a unified neural scheduling framework that adaptively selects high-value rollouts throughout training. Each rollout is treated as an arm whose reward is defined by the induced performance gain between consecutive optimization steps. The resulting scheduler supports both noise-aware intra-group selection and adaptive global reuse of historical rollouts within a single principled framework. We provide theoretical justification by deriving sublinear regret bounds and showing that enlarging the rollout buffer improves the achievable performance upper bound. Experiments on six mathematical reasoning benchmarks demonstrate consistent gains in performance and training efficiency across multiple RLVR optimization methods.

cs.LG↗
arXiv2026

$π$-Play: Multi-Agent Self-Play via Privileged Self-Distillation without External Data

Deep search agents have emerged as a promising paradigm for addressing complex information-seeking tasks, but their training remains challenging due to sparse rewards, weak credit assignment, and limited labeled data. Self-play offers a scalable route to reduce data dependence, but conventional self-play optimizes students only through sparse outcome rewards, leading to low learning efficiency. In this work, we observe that self-play naturally produces a question construction path (QCP) during task generation, an intermediate artifact that captures the reverse solution process. This reveals a new source of privileged information: self-play can provide high-quality privileged information for the self-distillation at low cost and at scale, without relying on human feedback or curated privileged information. Leveraging this insight, we propose Privileged Information Self-Play ($π$-Play), a novel multi-agent self-evolution framework combining self-play and self-distillation. In $π$-Play, an examiner generates tasks together with QCPs, and a teacher employs QCP as privileged context to densely supervise a student via self-distillation. This design transforms sparse-reward self-play into a dense-feedback co-evolution. Extensive experiments show that data-free $π$-Play surpasses fully supervised search agents and improves evolutionary efficiency by 2-3$\times$ over conventional self-play. Code is available at https://github.com/zhyaoch/pi-play.

cs.LG↗
arXiv2026

When Self-Belief Misleads: Active Label Acquisition for Reinforcement Learning with Verifiable Rewards

Large Language Models (LLMs) have achieved remarkable advancements in reasoning capabilities empowered by Reinforcement Learning with Verifiable Rewards (RLVR). Nonetheless, RLVR intrinsically relies on ground-truth labels for reward computation, the acquisition of which is often prohibitively expensive in real-world scenarios. While unsupervised RLVR paradigms attempt to circumvent this by training on pseudo-labels, they are notoriously susceptible to training collapse. Moreover, different samples often exhibit varying annotation values. In this paper, we propose Reinforcement Learning with Active Verifiable Rewards (RLAVR), which actively acquires ground-truth labels for a small set of selected samples and integrates them with pseudo-labels, thereby stabilizing training dynamics and improving performance under limited annotation budgets. To identify valuable samples, we propose the Corrective Advantage Gap (CAG) metric and analyze the sample-level supervision value. Building on this, we introduce Correction-Aware Reliability Estimation for RLAVR (CARE), which translates the oracle CAG criterion into a practical pre-query acquisition policy to substantially improve training stability. Extensive experiments across diverse domains, model families, and model scales demonstrate the effectiveness and generality of our approach. Our code is available at https://github.com/Lumina04/CARE.

cs.LG↗
arXiv2026

Exact controllability of the stochastic Maxwell equation: theory and numerical simulation

This article investigates the exact controllability of three-dimensional stochastic Maxwell equations, a coupled system comprising two stochastic partial differential equations. The research establishes the observability inequality for the backward stochastic Maxwell equations using the multiplier method, and subsequently, proves the exact controllability of the forward equations. The control acting on the diffusion term is found to be indispensable, since exact controllability is destroyed when this control is removed; it is further proved that the controllability result obtained in this paper is achieved with a minimal number of controls. Finally, a numerical algorithm combining a central difference for spatial discretization, a midpoint scheme for temporal discretization and Lagrange multiplier method is proposed, yielding numerical results that offer the control value and lead to deeper insights into the underlying theoretical framework.

math.OC↗
arXiv2026

Full-Spectrum Graph Neural Networks: Expressive and Scalable

It is well established that spectral graph neural networks (GNNs) can universally approximate node signals; however, their expressive power remains bounded by the 1-dimensional Weisfeiler-Lehman test, which is mirrored in their lack of universality for higher-order signals. To go beyond this bound, we propose the Full-Spectrum GNNs (FSpecGNNs), a second-order generalization of classical spectral GNNs. FSpecGNN advances spectral filtering from two perspectives: (1) it lifts signals from the node domain to the node-pair domain; and (2) it extends the univariate spectral filter over eigenvalues to a bivariate filter over eigenvalue pairs. We show that classical spectral GNNs arise as a diagonal special case of FSpecGNNs, and prove that FSpecGNNs can be at most as expressive as Local 2-GNN while universally approximating node-pair signals, the latter being particularly beneficial for heterophilic graph learning. Moreover, FSpecGNN admits scalable implementations that avoid explicit node-pair-level computations; combined with a low-rank approximation that reduces full-spectrum convolution to a combination of polynomial spectral filters, it enables learning on large graphs. Empirically, FSpecGNN validates the predicted expressivity and delivers strong performance on heterophilic benchmarks.

cs.LG↗
arXiv2026

FineVision: Open Data Is All You Need

The advancement of vision-language models (VLMs) is hampered by a fragmented landscape of inconsistent and contaminated public datasets. We introduce FineVision, a meticulously collected, curated, and unified corpus of 24 million samples - the largest open resource of its kind. We unify more than 200 sources into 185 subsets via a semi-automated, human-in-the-loop pipeline: automation performs bulk ingestion and schema mapping, while reviewers audit mappings and spot-check outputs to verify faithful consumption of annotations, appropriate formatting and diversity, and safety; issues trigger targeted fixes and re-runs. The workflow further applies rigorous de-duplication within and across sources and decontamination against 66 public benchmarks. FineVision also encompasses agentic/GUI tasks with a unified action space; reviewers validate schemas and inspect a sample of trajectories to confirm executable fidelity. Models trained on FineVision consistently outperform those trained on existing open mixtures across a broad evaluation suite, underscoring the benefits of scale, data hygiene, and balanced automation with human oversight. We release the corpus and curation tools to accelerate data-centric VLM research.

cs.CV↗