Searcharxiv⌕ Search

arXiv subjects

Xiaohan Wang

Publications and source records attributed to Xiaohan Wang.

At least 55 records · Page 3Linked to original sources
arXiv2026

Implicit Hierarchical GRPO: Decoupling Tool Invocation from Execution for Tool-Integrated Mathematical Reasoning

Large language models (LLMs) have increasingly leveraged tool invocation to enhance their reasoning capabilities. However, existing approaches typically tightly couple tool invocation with immediate execution. Such immediate tool interaction may disrupt the reasoning coherence of LLMs and constrain their expressivity, ultimately degrading reasoning performance. To this end, for the first time, we propose and formalize the problem of decoupling tool invocation from execution during reasoning, and introduce delayed execution with explicit control to enhance tool-integrated reasoning (TIR). Furthermore, we propose a hierarchical control framework and theoretically derive a surrogate loss that enables an implicitly hierarchical policy to learn behavior equivalent to that of an explicit hierarchical policy, leading to the proposed IH-GRPO algorithm. Extensive experiments on IH-GRPO achieve absolute improvements of 1.87\%, 2.16\%, and 2.53\% on Qwen3-1.7B, Qwen3-4B, and Qwen3-8B across six out-of-domain mathematical reasoning benchmarks over the strongest baseline method, while also yielding consistent performance gains in other domains. Our code is available at https://github.com/Lumina04/IH-GRPO-01.

cs.CL↗
arXiv2026

AMR-SD: Asymmetric Meta-Reflective Self-Distillation for Token-Level Credit Assignment

The alignment of Large Language Models (LLMs) for complex reasoning heavily relies on Reinforcement Learning with Verifiable Rewards (RLVR). However, standard algorithms like GRPO apply sequence-level rewards uniformly to all tokens, creating a severe credit-assignment bottleneck. While on-policy self-distillation attempts to resolve this by conditioning a self-teacher on privileged contexts, direct exposure to raw oracle solutions often induces over-conditioned teacher distributions, implicit answer leakage, and late-stage training collapse. To overcome these limitations, we propose Asymmetric Meta-Reflective Self-Distillation (AMR-SD). Instead of conditioning directly on raw reference traces, AMR-SD inserts a reflection bottleneck: it compresses diagnostic signals -- from verifier outcomes, peer rollouts, or reference feedback -- into concise, self-generated Socratic hints and critiques. Furthermore, we introduce Causal Information Gain (CIG) with an asymmetric, ReLU-gated threshold to translate these reflections into sparse, highly precise token-level advantage modulations. Combined with temporal annealing, this mechanism preserves the base environmental reward while filtering out distributional noise. Experiments across scientific, mathematical, and tool-use benchmarks demonstrate that AMR-SD significantly outperforms existing baselines, achieving robust long-horizon stability and successfully preventing late-stage collapse.

cs.AI↗
arXiv2026

SSL4RL: Revisiting Self-supervised Learning as Intrinsic Reward for Visual-Language Reasoning

Vision-language models (VLMs) have shown remarkable abilities by integrating large language models with visual inputs. However, they often fail to utilize visual evidence adequately, either depending on linguistic priors in vision-centric tasks or resorting to textual shortcuts during reasoning. Although reinforcement learning (RL) can align models with desired behaviors, its application to VLMs has been hindered by the lack of scalable and reliable reward mechanisms. To overcome this challenge, we propose SSL4RL, a novel framework that leverages self-supervised learning (SSL) tasks as a source of verifiable rewards for RL-based fine-tuning. Our approach reformulates SSL objectives-such as predicting image rotation or reconstructing masked patches-into dense, automatic reward signals, eliminating the need for human preference data or unreliable AI evaluators. Experiments show that SSL4RL substantially improves performance on both vision-centric and vision-language reasoning benchmarks. Furthermore, through systematic ablations, we identify key factors-such as task difficulty, model scale, and semantic alignment with the target domain-that influence the effectiveness of SSL4RL tasks, offering new design principles for future work. We also demonstrate the framework's generality by applying it to graph learning, where it yields significant gains. SSL4RL establishes a versatile and effective paradigm for aligning multimodal models using verifiable, self-supervised objectives.

cs.CV↗
arXiv2026

RecRM-Bench: Benchmarking Multidimensional Reward Modeling for Agentic Recommender Systems

The integration of Large Language Model (LLM) agents is transforming recommender systems from simple query-item matching towards deeply personalized and interactive recommendations. Reinforcement Learning (RL) provides an essential framework for the optimization of these agents in recommendation tasks. However, current methodologies remain limited by a reliance on single dimensional outcome-based rewards that focus exclusively on final user interactions, overlooking critical intermediate capabilities, such as instruction following and complex intent understanding. Despite the necessity for designing multi-dimensional reward, the field lacks a standardized benchmark to facilitate this development. To bridge this gap, we introduce RecRM-Bench, the largest and most comprehensive benchmark to date for agentic recommender systems. It comprises over 1 million structured entries across four core evaluation dimensions: instruction following, factual consistency, query-item relevance, and fine-grained user behavior prediction. By supporting comprehensive assessment from syntactic compliance to complex intent grounding and preference modeling, RecRM-Bench provides a foundational dataset for training sophisticated reward models. Furthermore, we propose a systematic framework for the construction of multi-dimensional reward models and the integration of a hybrid reward function, establishing a robust foundation for developing reliable and highly capable agentic recommender systems. The complete RecRM-Bench dataset is publicly available at https://huggingface.co/datasets/wwzeng/RecRM-Bench.

cs.IR↗
arXiv2026

Concise Geometric Description as a Bridge: Unleashing the Potential of LLM for Plane Geometry Problem Solving

Plane Geometry Problem Solving (PGPS) is a multimodal reasoning task that aims to solve a plane geometric problem based on a geometric diagram and problem textual descriptions. Although Large Language Models (LLMs) possess strong reasoning skills, their direct application to PGPS is hindered by their inability to process visual diagrams. Existing works typically fine-tune Multimodal LLMs (MLLMs) end-to-end on large-scale PGPS data to enhance visual understanding and reasoning simultaneously. However, such joint optimization may compromise base LLMs' inherent reasoning capability. In this work, we observe that LLM itself is potentially a powerful PGPS solver when appropriately formulating visual information as textual descriptions. We propose to train a MLLM Interpreter to generate geometric descriptions for the visual diagram, and an off-the-shelf LLM is utilized to perform reasoning. Specifically, we choose Conditional Declaration Language (CDL) as the geometric description as its conciseness eases the MLLM Interpreter training. The MLLM Interpreter is fine-tuned via CoT (Chain-of-Thought)-augmented SFT followed by GRPO to generate CDL. Instead of using a conventional solution-based reward that compares the reasoning result with the ground-truth answer, we design CDL matching rewards to facilitate more effective GRPO training, which provides more direct and denser guidance for CDL generation. To support training, we construct a new dataset, Formalgeo7k-Rec-CoT, by manually reviewing Formalgeo7k v2 and incorporating CoT annotations. Extensive experiments on Formalgeo7k-Rec-CoT, Unigeo, and MathVista show our method (finetuned on only 5.5k data) performs favorably against leading open-source and closed-source MLLMs.

cs.AI↗
arXiv2026

ResRL: Boosting LLM Reasoning via Negative Sample Projection Residual Reinforcement Learning

Reinforcement Learning with Verifiable Rewards (RLVR) enhances reasoning of Large Language Models (LLMs) but usually exhibits limited generation diversity due to the over-incentivization of positive rewards. Although methods like Negative Sample Reinforcement (NSR) mitigate this issue by upweighting penalty from negative samples, they may suppress the semantic distributions shared between positive and negative responses. To boost reasoning ability without losing diversity, this paper proposes negative sample projection Residual Reinforcement Learning (ResRL) that decouples similar semantic distributions among positive and negative responses. We theoretically link Lazy Likelihood Displacement (LLD) to negative-positive head-gradient interference and derive a single-forward proxy that upper-bounds representation alignment to guide conservative advantage reweighting. ResRL then projects negative-token hidden representations onto an SVD-based low-rank positive subspace and uses projection residuals to modulate negative gradients, improving reasoning while preserving diversity and outperforming strong baselines on average across twelve benchmarks spanning Mathematics, Code, Agent Tasks, and Function Calling. Notably, ResRL surpasses NSR on mathematical reasoning by 9.4\% in Avg@16 and 7.0\% in Pass@128. Code is available at https://github.com/1229095296/ResRL.git.

cs.LG↗
arXiv2026

Uncovering Entity Identity Confusion in Multimodal Knowledge Editing

Multimodal knowledge editing (MKE) aims to correct the internal knowledge of large vision-language models after deployment, yet the behavioral patterns of post-edit models remain underexplored. In this paper, we identify a systemic failure mode in edited models, termed Entity Identity Confusion (EIC): edited models exhibit an absurd behavior where text-only queries about the original entity's identity unexpectedly return information about the new entity. To rigorously investigate EIC, we construct EC-Bench, a diagnostic benchmark that directly probes how image-entity bindings shift before and after editing. Our analysis reveals that EIC stems from existing methods failing to distinguish between Image-Entity (I-E) binding and Entity-Entity (E-E) relational knowledge in the model, causing models to overfit E-E associations as a shortcut: the image is still perceived as the original entity, with the new entity's name serving only as a spurious identity label. We further explore potential mitigation strategies, showing that constraining edits to the model's I-E processing stage encourages edits to act more faithfully on I-E binding, thereby substantially reducing EIC. Based on these findings, we discuss principled desiderata for faithful MKE and provide methodological guidance for future research.

cs.CL↗
arXiv2026

Constraint-Guided Multi-Agent Decompilation for Executable Binary Recovery

Decompilation -- recovering source code from compiled binaries -- is essential for security analysis, malware reverse engineering, and legacy software maintenance. However, existing decompilers produce code that often fails to compile or execute correctly, limiting their practical utility. We present a multi-agent framework that transforms decompiled code into re-executable source through Multi-level Constraint-Guided Decompilation (MCGD). Our approach employs a hierarchical validation pipeline with three constraint levels: (1) syntactic correctness via parsing, (2) compilability via GCC, and (3) behavioral equivalence via LLM-generated test cases. When validation fails, specialized LLM agents iteratively refine the code using structured error feedback. We evaluate our framework on 1,641 real-world binaries from ExeBench across three decompilers (RetDec, Ghidra, and Angr). Our framework achieves 84-97% re-executability, improving baseline decompiler output by 28-89 percentage points. In comparison with state-of-the-art LLM-based decompilation methods using the same GPT-4o backbone, our approach (84.1%) outperforms LLM4Decompile (80.3%), SK2Decompile (73.9%), and SALT4Decompile (61.8%). Our ablation study reveals that execution-based validation is critical: compile-only approaches achieve 0% behavioral correctness despite 91-99% compilation rates. The system converges efficiently, with 90%+ binaries reaching correctness within 2 iterations at an average cost of $0.03-0.05 per binary. Our results demonstrate that constraint-guided agentic refinement can bridge the gap between raw decompiler output and practically useful source code.

cs.SE↗
arXiv2026

V-GRPO: Online Reinforcement Learning for Denoising Generative Models Is Easier than You Think

Aligning denoising generative models with human preferences or verifiable rewards remains a key challenge. While policy-gradient online reinforcement learning (RL) offers a principled post-training framework, its direct application is hindered by the intractable likelihoods of these models. Prior work therefore either optimizes an induced Markov decision process (MDP) over sampling trajectories, which is stable but inefficient, or uses likelihood surrogates based on the diffusion evidence lower bound (ELBO), which have so far underperformed on visual generation. Our key insight is that the ELBO-based approach can, in fact, be made both stable and efficient. By reducing surrogate variance and controlling gradient steps, we show that this approach can beat MDP-based methods. To this end, we introduce Variational GRPO (V-GRPO), a method that integrates ELBO-based surrogates with the Group Relative Policy Optimization (GRPO) algorithm, alongside a set of simple yet essential techniques. Our method is easy to implement, aligns with pretraining objectives, and avoids the limitations of MDP-based methods. V-GRPO achieves state-of-the-art performance in text-to-image synthesis, while delivering a $2\times$ speedup over MixGRPO and a $3\times$ speedup over DiffusionNFT.

cs.LG↗
arXiv2026

Petabit-per-second Random Number Generation

Physical random number generators based on chaotic microcombs, with their complex nonlinear dynamics and multi-channel parallel capability, have attracted considerable research attention. However, key technical challenges for chaotic microcombs are the high correlation between symmetric teeth and the low bandwidth of single-channel teeth, which seriously affect the speed and scalability of random number generation. We experimentally demonstrate a petabit-per-second (Pbit/s) parallel random number generation system based on intensity chaotic modulation and Rayleigh scattering. Through intensity modulation, the effective bandwidth of the single-channel entropy source is increased from 440MHz to 27.6GHz. Crucially, Rayleigh scattering further contributes through the random superposition of backscattered light, which introduces unpredictable fluctuations in intensity, phase, and polarization. This randomness suppresses inter-channel correlation among parallel entropy sources to ~0.02, ensuring their orthogonality. Moreover, by employing polarization-diverse coherent detection on a single-channel, four new low correlated sub-channels are extracted: X-/Y- intensity and phase. We achieve a single-channel bit rate of 14.336 Tbit/s and a total bit rate of 1.032 Pbit/s (over 72 parallel channels) with offline post-processing, representing the highest post-processing record reported in both the single-channel and the total system. Moreover, our scheme based on a single chaotic microcomb and fiber scattering link show fundamentally scalable. The total bit rate can be significantly pushed beyond the Pbit/s level by further expanding the usable comb channel and/or by deploying multiple fiber scattering links in parallel, paving a practical path toward higher throughput regimes.

physics.optics↗
arXiv2026

AutoSearch: Adaptive Search Depth for Efficient Agentic RAG via Reinforcement Learning

Agentic retrieval-augmented generation (RAG) systems enable large language models (LLMs) to solve complex tasks through multi-step interaction with external retrieval tools. However, such multi-step interaction often involves redundant search steps, incurring substantial computational cost and latency. Prior work limits search depth (i.e., the number of search steps) to reduce cost, but this often leads to underexploration of complex questions. To address this, we first investigate how search depth affects accuracy and find a minimal sufficient search depth that defines an accuracy-efficiency trade-off, jointly determined by question complexity and the agent's capability. Furthermore, we propose AutoSearch, a reinforcement learning (RL) framework that evaluates each search step via self-generated intermediate answers. By a self-answering mechanism, AutoSearch identifies the minimal sufficient search depth and promotes efficient search by rewarding its attainment while penalizing over-searching. In addition, reward mechanisms are introduced to stabilize search behavior and improve answer quality on complex questions. Extensive experiments on multiple benchmarks show that AutoSearch achieves a superior accuracy-efficiency trade-off, alleviating over-searching while preserving search quality.

cs.AI↗
arXiv2026

Context-Guided Decompilation: A Step Towards Re-executability

Binary decompilation plays an important role in software security analysis, reverse engineering, and malware understanding when source code is unavailable. However, existing decompilation techniques often fail to produce source code that can be successfully recompiled and re-executed, particularly for optimized binaries. Recent advances in large language models (LLMs) have enabled neural approaches to decompilation, but the generated code is typically only semantically plausible rather than truly executable, limiting their practical reliability. These shortcomings arise from compiler optimizations and the loss of semantic cues in compiled code, which LLMs struggle to recover without contextual guidance. To address this challenge, we propose ICL4Decomp, a hybrid decompilation framework that leverages in-context learning (ICL) to guide LLMs toward generating re-executable source code. We evaluate our method across multiple datasets, optimization levels, and compilers, demonstrating around 40\% improvement in re-executability over state-of-the-art decompilation methods while maintaining robustness.

cs.SE↗
arXiv2026

UHD Low-Light Image Enhancement via Real-Time Enhancement Methods with Clifford Information Fusion

Considering efficiency, ultra-high-definition (UHD) low-light image restoration is extremely challenging. Existing methods based on Transformer architectures or high-dimensional complex convolutional neural networks often suffer from the "memory wall" bottleneck, failing to achieve millisecond-level inference on edge devices. To address this issue, we propose a novel real-time UHD low-light enhancement network based on geometric feature fusion using Clifford algebra in 2D Euclidean space. First, we construct a four-layer feature pyramid with gradually increasing resolution, which decomposes input images into low-frequency and high-frequency structural components via a Gaussian blur kernel, and adopts a lightweight U-Net based on depthwise separable convolution for dual-branch feature extraction. Second, to resolve structural information loss and artifacts from traditional high-low frequency feature fusion, we introduce spatially aware Clifford algebra, which maps feature tensors to a multivector space (scalars, vectors, bivectors) and uses Clifford similarity to aggregate features while suppressing noise and preserving textures. In the reconstruction stage, the network outputs adaptive Gamma and Gain maps, which perform physically constrained non-linear brightness adjustment via Retinex theory. Integrated with FP16 mixed-precision computation and dynamic operator fusion, our method achieves millisecond-level inference for 4K/8K images on a single consumer-grade device, while outperforming state-of-the-art (SOTA) models on several restoration metrics.

eess.IV↗
arXiv2026

GLANCE: A Global-Local Coordination Multi-Agent Framework for Music-Grounded Non-Linear Video Editing

Music-grounded mashup video creation is a challenging form of video non-linear editing, where a system must compose a coherent timeline from large collections of source videos while aligning with music rhythm, user intent, story completeness, and long-range structural constraints. Existing approaches typically rely on fixed pipelines or simplified retrieval-and-concatenation paradigms, limiting their ability to adapt to diverse prompts and heterogeneous source materials. In this paper, we present GLANCE, a global-local coordination multi-agent framework for music-grounded nonlinear video editing. GLANCE adopts a bi-loop architecture for better editing practice: an outer loop performs long-horizon planning and task-graph construction, and an inner loop adopts the "Observe-Think-Act-Verify" flow for segment-wise editing tasks and their refinements. To address the cross-segment and global conflict emerging after subtimelines composition, we introduce a dedicated global-local coordination mechanism with both preventive and corrective components, which includes a novelly designed context controller, conflict region decomposition module, and a bottom-up dynamic negotiation mechanism. To support rigorous evaluation, we construct MVEBench, a new benchmark that factorizes editing difficulty along task type, prompt specificity, and music length, and propose an agent-as-a-judge evaluation framework for scalable multi-dimensional assessment. Experimental results show that GLANCE consistently outperforms prior research baselines and open-source product baselines under the same backbone models. With GPT-4o-mini as the backbone, GLANCE improves over the strongest baseline by 33.2% and 15.6% on two task settings, respectively. Human evaluation further confirms the quality of the generated videos and validates the effectiveness of the proposed evaluation framework.

cs.MA↗
arXiv2026

Training Multi-Image Vision Agents via End2End Reinforcement Learning

Recent VLM-based agents aim to replicate OpenAI O3's "thinking with images" via tool use, yet most open-source methods restrict inputs to a single image, limiting their applicability to real-world multi-image QA tasks. To address this gap, we propose IMAgent, an open-source visual agent trained with end-to-end reinforcement learning for fine-grained single/multi-image reasoning. During inference, VLMs tend to gradually neglect visual inputs; to mitigate this issue, we design two dedicated tools for visual reflection and verification, enabling the model to actively refocus attention on image content. Beyond that, we, for the first time, reveal how tool usage enhances agent performance from an attention perspective. Equipped with a carefully designed two-layer motion trajectory masking strategy and tool-use reward gain, IMAgent acquires an effective tool-use paradigm through pure reinforcement learning, eliminating the need for costly supervised fine-tuning data. To further unleash the inherent tool-usage potential of the base VLM and fill data gaps, we construct a challenging, visually enriched multi-image QA dataset via multi-agent system. Extensive experiments validate that IMAgent achieves SOTA performance across mainstream single and multi-image benchmarks, and our in-depth analysis offers actionable insights for the community. Code and data will be released soon.

cs.CV↗
arXiv2026

GMPilot: An Expert AI Agent For FDA cGMP Compliance

The pharmaceutical industry is facing challenges with quality management such as high costs of compliance, slow responses and disjointed knowledge. This paper presents GMPilot, a domain-specific AI agent that is designed to support FDA cGMP compliance. GMPilot is based on a curated knowledge base of regulations and historical inspection observations and uses Retrieval-Augmented Generation (RAG) and Reasoning-Acting (ReAct) frameworks to provide real-time and traceable decision support to the quality professionals. In a simulated inspection scenario, GMPilot shows how it can improve the responsiveness and professionalism of quality professionals by providing structured knowledge retrieval and verifiable regulatory and case-based support. Although GMPilot lacks in the aspect of regulatory scope and model interpretability, it is a viable avenue of improving quality management decision-making in the pharmaceutical sector using intelligent approaches and an example of specialized application of AI in highly regulated sectors.

cs.AI↗
arXiv2026

Fine-tuning MLLMs Without Forgetting Is Easier Than You Think

The paper demonstrate that simple adjustments of the fine-tuning recipes of multimodal large language models (MLLM) are sufficient to mitigate catastrophic forgetting. On visual question answering, we design a 2x2 experimental framework to assess model performance across in-distribution and out-of-distribution image and text inputs. Our results show that appropriate regularization, such as constraining the number of trainable parameters or adopting a low learning rate, effectively prevents forgetting when dealing with out-of-distribution images. However, we uncover a distinct form of forgetting in settings with in-distribution images and out-of-distribution text. We attribute this forgetting as task-specific overfitting and address this issue by introducing a data-hybrid training strategy that combines datasets and tasks. Finally, we demonstrate that this approach naturally extends to continual learning, outperforming existing methods with complex auxiliary mechanisms. In general, our findings challenge the prevailing assumptions by highlighting the inherent robustness of MLLMs and providing practical guidelines for adapting them while preserving their general capabilities.

cs.CV↗
arXiv2026

CDRRM: Contrast-Driven Rubric Generation for Reliable and Interpretable Reward Modeling

Reward modeling is essential for aligning Large Language Models(LLMs) with human preferences, yet conventional reward models suffer from poor interpretability and heavy reliance on costly expert annotations. While recent rubric-based approaches enhance evaluation transparency, they lack systematic quality control, yielding noisy and redundant criteria, failing to mitigate persistent biases (e.g., verbosity, position) in LLM evaluators, and creating a scalability-reliability trade-off. To address these limitations, we propose CDRRM (Contrast-Driven Rubric Reward Model), a framework built on a novel Contrast-then-Synthesis paradigm for high-quality rubric generation and guided preference judgment. CDRRM first conducts multi-dimensional contrastive profiling on preference pairs to identify causal discriminative factors, then synthesizes these insights into compact, context-aware rubrics to guide preference judg- ments. Extensive experiments on three authoritative benchmarks (RewardBench, RMBench, RMB) demonstrate that CDRRM achieves state-of-the-art performance across diverse domains and effectively mitigates aforementioned evaluation biases. Notably, our approach delivers exceptional data efficiency: training the rubric generator on only 3k high-quality samples empowers a frozen pre-trained judge model to outperform fully fine-tuned baselines. This work offers a scalable, interpretable, and data-efficient path for reward modeling.

cs.AI↗