arXiv subjects

Peng Wang

Publications and source records attributed to Peng Wang.

At least 181 records · Page 10Linked to original sources

arXiv2026

Reasoning-Driven Anomaly Detection and Localization with Image-Level Supervision

Multimodal large language models (MLLMs) have recently demonstrated remarkable reasoning and perceptual abilities for anomaly detection. However, most approaches remain confined to image-level anomaly detection and textual reasoning, while pixel-level localization still relies on external vision modules and dense annotations. In this work, we activate the intrinsic reasoning potential of MLLMs to perform anomaly detection, pixel-level localization, and interpretable reasoning solely from image-level supervision, without any auxiliary components or pixel-wise labels. Specifically, we propose Reasoning-Driven Anomaly Localization (ReAL), which extracts anomaly-related tokens from the autoregressive reasoning process and aggregates their attention responses to produce pixel-level anomaly maps. We further introduce a Consistency-Guided Reasoning Optimization (CGRO) module that leverages reinforcement learning to align reasoning tokens with visual attentions, resulting in more coherent reasoning and accurate anomaly localization. Extensive experiments on four public benchmarks demonstrate that our method significantly improves anomaly detection, localization, and interpretability. Remarkably, despite relying solely on image-level supervision, our approach achieves performance competitive with MLLM-based methods trained under dense pixel-level supervision. Code is available at https://github.com/YizhouJin313/ReADL.

cs.CV↗

arXiv2026

Understanding and Mitigating Hallucinations in Multimodal Chain-of-Thought Models

Multimodal Chain-of-Thought (MCoT) models have demonstrated impressive capability in complex visual reasoning tasks. Unfortunately, recent studies reveal that they suffer from severe hallucination problems due to diminished visual attention during the generation process. However, visual attention decay is a well-studied problem in Large Vision-Language Models (LVLMs). Considering the fundamental differences in reasoning processes between MCoT models and traditional LVLMs, we raise a basic question: Whether MCoT models have unique causes of hallucinations? To answer this question, we systematically investigate the hallucination patterns of MCoT models and find that fabricated texts are primarily generated in associative reasoning steps, which we term divergent thinking. Leveraging these insights, we introduce a simple yet effective strategy that can effectively localize divergent thinking steps and intervene in the decoding process to mitigate hallucinations. Extensive experiments show that our method outperforms existing methods by a large margin. More importantly, our proposed method can be conveniently integrated with other hallucination mitigation methods and further boost their performance. The code is publicly available at https://github.com/ASGO-MM/MCoT-hallucination.

cs.CV↗

arXiv2026

Immersion in the GitHub Universe: Scaling Coding Agents to Mastery

Achieving mastery in real world software engineering tasks is fundamentally bottlenecked by the scarcity of large scale, high quality training data. Scaling such data has been limited by the complexity of environment setup, unit test generation, and problem statement curation. In this paper, we propose ScaleSWE, an automated, sandboxed multi agent workflow designed to construct high quality SWE data at scale. The system coordinates three specialized agents for environment setup, test creation, and problem description synthesis to process 6 million pull requests across 5200 repositories, producing Scale SWE Data: 100k verified SWE instances, the largest such dataset to date. It substantially surpasses existing real world datasets in repository diversity and reflects realistic task complexity. We further demonstrate the dataset utility for training by distilling 71498 high quality trajectories and finetuning Qwen30BA3BInstruct to produce ScaleSWE Agent. Our agent achieves a 64 resolve rate on SWE Bench Verified a nearly three fold improvement over the base model. ScaleSWE provides a scalable, reproducible approach for data construction to advance LLM based software engineering. Scale SWE will be publicly available.

cs.SE↗

arXiv2026

GO-Renderer: Generative Object Rendering with 3D-aware Controllable Video Diffusion Models

Reconstructing a renderable 3D model from images is a useful but challenging task. Recent feedforward 3D reconstruction methods have demonstrated remarkable success in efficiently recovering geometry, but still cannot accurately model the complex appearances of these 3D reconstructed models. Recent diffusion-based generative models can synthesize realistic images or videos of an object using reference images without explicitly modeling its appearance, which provides a promising direction for object rendering, but lacks accurate control over the viewpoints. In this paper, we propose GO-Renderer, a unified framework integrating the reconstructed 3D proxies to guide the video generative models to achieve high-quality object rendering on arbitrary viewpoints under arbitrary lighting conditions. Our method not only enjoys the accurate viewpoint control using the reconstructed 3D proxy but also enables high-quality rendering in different lighting environments using diffusion generative models without explicitly modeling complex materials and lighting. Extensive experiments demonstrate that GO-Renderer achieves state-of-the-art performance across the object rendering tasks, including synthesizing images on new viewpoints, rendering the objects in a novel lighting environment, and inserting an object into an existing video.

cs.CV↗

arXiv2026

SQL-Commenter: Aligning Large Language Models for SQL Comment Generation with Direct Preference Optimization

SQL query comprehension is a significant challenge due to complex syntax, diverse join types, and deep nesting. Many queries lack adequate comments, severely hindering code readability, maintainability, and knowledge transfer. Automated SQL comment generation faces two main challenges: limited datasets that inadequately represent complex real-world queries, and Large Language Models' (LLMs) insufficient understanding of SQL-specific semantics. Our empirical analysis shows that even after continual pre-training and supervised fine-tuning, LLMs struggle with complex SQL semantics, yielding inaccurate comments. To address this, we propose SQL-Commenter, an advanced method based on LLaMA-3.1-8B. We first construct a comprehensive dataset of complex SQL queries with expert-verified comments. Next, we perform continual pre-training on a large SQL corpus to enhance the LLM's syntax and semantic understanding, followed by supervised fine-tuning. Finally, we introduce Direct Preference Optimization (DPO) using human feedback. SQL-Commenter utilizes a preference-based loss function to favor preferred outputs, enhancing fine-grained semantic learning and context-dependent quality assessment. Evaluated on the Spider and Bird benchmarks, SQL-Commenter significantly outperforms state-of-the-art baselines. On average, it surpasses the strongest baseline (Qwen3-14B) by 9.29, 4.99, and 13.23 percentage points on BLEU-4, METEOR, and ROUGE-L, respectively. Moreover, human evaluation demonstrates the superior quality of comments generated by SQL-Commenter in terms of correctness, completeness, and naturalness.

cs.SE↗

arXiv2026

Linearly Separable Features in Shallow Nonlinear Networks: Width Scales Polynomially with Intrinsic Data Dimension

Deep neural networks have attained remarkable success across diverse classification tasks. Recent empirical studies have shown that deep networks learn features that are linearly separable across classes. However, these findings often lack rigorous justifications, even under relatively simple settings. In this work, we address this gap by examining the linear separation capabilities of shallow nonlinear networks. Specifically, inspired by the low intrinsic dimensionality of image data, we model inputs as a union of low-dimensional subspaces (UoS) and demonstrate that a single nonlinear layer can transform such data into linearly separable sets. Theoretically, we show that this transformation occurs with high probability when using random weights and quadratic activations. Notably, we prove this can be achieved when the network width scales polynomially with the intrinsic dimension of the data rather than the ambient dimension. Experimental results corroborate these theoretical findings and demonstrate that similar linear separation properties hold in practical scenarios beyond our analytical scope. This work bridges the gap between empirical observations and theoretical understanding of the separation capacity of nonlinear networks, offering deeper insights into model interpretability and generalization.

cs.LG↗

arXiv2026

Long-LRM++: Preserving Fine Details in Feed-Forward Wide-Coverage Reconstruction

Recent advances in generalizable Gaussian splatting (GS) have enabled feed-forward reconstruction of scenes from tens of input views. Long-LRM notably scales this paradigm to 32 input images at $950\times540$ resolution, achieving 360° scene-level reconstruction in a single forward pass. However, directly predicting millions of Gaussian parameters at once remains highly error-sensitive: small inaccuracies in positions or other attributes lead to noticeable blurring, particularly in fine structures such as text. In parallel, implicit representation methods such as LVSM and LaCT have demonstrated significantly higher rendering fidelity by compressing scene information into model weights rather than explicit Gaussians, and decoding RGB frames using the full transformer or TTT backbone. However, this computationally intensive decompression process for every rendered frame makes real-time rendering infeasible. These observations raise key questions: Is the deep, sequential "decompression" process necessary? Can we retain the benefits of implicit representations while enabling real-time performance? We address these questions with Long-LRM++, a model that adopts a semi-explicit scene representation combined with a lightweight decoder. Long-LRM++ matches the rendering quality of LaCT on DL3DV while achieving real-time 14 FPS rendering on an A100 GPU, overcoming the speed limitations of prior implicit methods. Our design also scales to 64 input views at the $950\times540$ resolution, demonstrating strong generalization to increased input lengths. Additionally, Long-LRM++ delivers superior novel-view depth prediction on ScanNetv2 compared to direct depth rendering from Gaussians. Extensive ablation studies validate the effectiveness of each component in the proposed framework.

cs.CV↗

arXiv2026

Deep Tabular Representation Corrector

Tabular data have been playing a mostly important role in diverse real-world fields, such as healthcare, engineering, finance, etc. The recent success of deep learning has fostered many deep networks (e.g., Transformer, ResNet) based tabular learning methods. Generally, existing deep tabular machine learning methods are along with the two paradigms, i.e., in-learning and pre-learning. In-learning methods need to train networks from scratch or impose extra constraints to regulate the representations which nonetheless train multiple tasks simultaneously and make learning more difficult, while pre-learning methods design several pretext tasks for pre-training and then conduct task-specific fine-tuning, which however need much extra training effort with prior knowledge. In this paper, we introduce a novel deep Tabular Representation Corrector, TRC, to enhance any trained deep tabular model's representations without altering its parameters in a model-agnostic manner. Specifically, targeting the representation shift and representation redundancy that hinder prediction, we propose two tasks, i.e., (i) Tabular Representation Re-estimation, that involves training a shift estimator to calculate the inherent shift of tabular representations to subsequently mitigate it, thereby re-estimating the representations and (ii) Tabular Space Mapping, that transforms the above re-estimated representations into a light-embedding vector space via a coordinate estimator while preserves crucial predictive information to minimize redundancy. The two tasks jointly enhance the representations of deep tabular models without touching on the original models thus enjoying high efficiency. Finally, we conduct extensive experiments on state-of-the-art deep tabular machine learning models coupled with TRC on various tabular benchmarks which have shown consistent superiority.

cs.LG↗

arXiv2026

Exploring Low-Dimensional Subspaces in Diffusion Models for Controllable Image Editing

Recently, diffusion models have emerged as a powerful class of generative models. Despite their success, there is still limited understanding of their semantic spaces. This makes it challenging to achieve precise and disentangled image generation without additional training, especially in an unsupervised way. In this work, we improve the understanding of their semantic spaces from intriguing observations: among a certain range of noise levels, (1) the learned posterior mean predictor (PMP) in the diffusion model is locally linear, and (2) the singular vectors of its Jacobian lie in low-dimensional semantic subspaces. We provide a solid theoretical basis to justify the linearity and low-rankness in the PMP. These insights allow us to propose an unsupervised, single-step, training-free LOw-rank COntrollable image editing (LOCO Edit) method for precise local editing in diffusion models. LOCO Edit identified editing directions with nice properties: homogeneity, transferability, composability, and linearity. These properties of LOCO Edit benefit greatly from the low-dimensional semantic subspace. Our method can further be extended to unsupervised or text-supervised editing in various text-to-image diffusion models (T-LOCO Edit). Finally, extensive empirical experiments demonstrate the effectiveness and efficiency of LOCO Edit.

cs.CV↗

arXiv2026

An Ultra-Low Power and Fast Ising Machine using Voltage-Controlled Magnetoresistive Random Access Memory

Physics-inspired computing paradigms, such as Ising machines, are emerging as promising hardware alternatives to traditional von Neumann architectures for tackling computationally intensive combinatorial optimization problems (COPs). While quantum, optical, and electronic devices have garnered significant attention for their potential in realizing Ising machines, their translation into practical systems for industry-relevant applications remains challenging, with each approach facing specific limitations in power consumption and speed. To address this challenge, we report the first chip-level spintronic Ising machine using voltage-controlled magnetoresistive random access memory. The core of our design leverages magnetic tunnel junctions (MTJs) driven by the voltage-controlled magnetic anisotropy effect to realize the probabilistic update of Ising spins through a new mechanism. It enables a latency below 1 ns and an energy consumption under 40 fJ per spin update, achieving a 1000-times improvement over previous current-driven MTJ-based implementations. We map two real-world COPs in electronic design automation-global routing and layer assignment-onto the Ising model and demonstrate high-quality results with an energy efficiency of 25000 solutions per second per watt. This outperforms state-of-the-art quantum and graphics processing units by six and seven orders of magnitude, respectively. These results establish voltage-controlled spintronics as a compelling route towards next-generation physics-inspired machine intelligence, offering a paradigm for ultra-low-power, high-speed, and scalable computation.

physics.app-ph↗

arXiv2026

Efficient and Interpretable Multi-Agent LLM Routing via Ant Colony Optimization

Large Language Model (LLM)-driven Multi-Agent Systems (MAS) have demonstrated strong capability in complex reasoning and tool use, and heterogeneous agent pools further broaden the quality--cost trade-off space. Despite these advances, real-world deployment is often constrained by high inference cost, latency, and limited transparency, which hinders scalable and efficient routing. Existing routing strategies typically rely on expensive LLM-based selectors or static policies, and offer limited controllability for semantic-aware routing under dynamic loads and mixed intents, often resulting in unstable performance and inefficient resource utilization. To address these limitations, we propose AMRO-S, an efficient and interpretable routing framework for Multi-Agent Systems (MAS). AMRO-S models MAS routing as a semantic-conditioned path selection problem, enhancing routing performance through three key mechanisms: First, it leverages a supervised fine-tuned (SFT) small language model for intent inference, providing a low-overhead semantic interface for each query; second, it decomposes routing memory into task-specific pheromone specialists, reducing cross-task interference and optimizing path selection under mixed workloads; finally, it employs a quality-gated asynchronous update mechanism to decouple inference from learning, optimizing routing without increasing latency. Extensive experiments on five public benchmarks and high-concurrency stress tests demonstrate that AMRO-S consistently improves the quality--cost trade-off over strong routing baselines, while providing traceable routing evidence through structured pheromone patterns.

cs.AI↗

arXiv2026

Language-Guided Token Compression with Reinforcement Learning in Large Vision-Language Models

Large Vision-Language Models (LVLMs) incur substantial inference costs due to the processing of a vast number of visual tokens. Existing methods typically struggle to model progressive visual token reduction as a multi-step decision process with sequential dependencies and often rely on hand-engineered scoring rules that lack adaptive optimization for complex reasoning trajectories. To overcome these limitations, we propose TPRL, a reinforcement learning framework that learns adaptive pruning trajectories through language-guided sequential optimization tied directly to end-task performance. We formulate visual token pruning as a sequential decision process with explicit state transitions and employ a self-supervised autoencoder to compress visual tokens into a compact state representation for efficient policy learning. The pruning policy is initialized through learning from demonstrations and subsequently fine-tuned using Proximal Policy Optimization (PPO) to jointly optimize task accuracy and computational efficiency. Our experimental results demonstrate that TPRL removes up to 66.7\% of visual tokens and achieves up to a 54.2\% reduction in FLOPs during inference while maintaining a near-lossless average accuracy drop of only 0.7\%. Code is released at \href{https://github.com/MagicVicCoder/TPRL}{\textcolor{mypink}{https://github.com/MagicVicCoder/TPRL}}.

cs.CV↗

arXiv2026

Beyond Static: Related Questions Retrieval Through Conversations in Community Question Answering

In community question answering (cQA) platforms like Stack Overflow, related question retrieval is recognized as a fundamental task that allows users to retrieve related questions to answer user queries automatically. Although many traditional approaches have been proposed for investigating this research field, they mostly rely on static approaches and neglect the interaction property. We argue that the conversational way can well distinguish the fine-grained representations of questions and has great potential to improve the performance of question retrieval. In this paper, we propose a related question retrieval model through conversations, called TeCQR, to locate related questions in cQA. Specifically, we build conversations by utilizing tag-enhanced clarifying questions (CQs). In addition, we design a noise tolerance model that evaluates the semantic similarity between questions and tags, enabling the model to effectively handle noisy feedback. Moreover, the tag-enhanced two-stage offline training is proposed to fully exploit the mutual relationships among user queries, questions, and tags to learn their fine-grained representations. Based on the learned representations and contextual conversations, TeCQR incorporates conversational feedback by learning to ask tag-enhanced clarifying questions to retrieve related questions more effectively. Experimental results demonstrate that our model significantly outperforms state-of-the-art baselines.

cs.IR↗

arXiv2026

Prolate-oblate shape competition and impact on charge radii in Bk isotopes

The nuclear charge radius provides a fundamental probe of nuclear structure, yet experimental data remain rare in the actinide region. Using the deformed relativistic Hartree-Bogoliubov theory in continuum (DRHBc) with the PC-PK1 functional, we carry out a systematic investigation of prolate-oblate shape competition in odd-$A$ Bk isotopes. Deformation is found to play an important role in the description of charge radii $r_c$ by extending the density distribution. Notably, $r_c$ exhibits a distinct shape dependence: for a given absolute quadrupole deformation $|β_2|$, oblate shapes yield larger charge radii than their prolate counterparts in well-deformed nuclei near the mid-shell region, where the empirical formula $r_c(β_2) = \left(1 + \frac{5}{4π}|β_2|^2\right) r_c(0)$ fails to capture the observed behavior. This enhancement is attributed to a central depression (or ``bubble" structure) in the proton density, which microscopically originates from the non-occupation of the spherical $3s_{1/2}$($Ω=1/2$) orbital in oblate minima. These findings establish a clear microscopic connection between nuclear shape, single-particle occupancy, and nuclear size.

nucl-th↗

arXiv2026

Coexistence of Spectrally Stable and Unstable Modes in Black Hole Ringdowns

Recent studies have shown that a secondary potential barrier, forming a potential well outside the event horizon, can destabilize the Quasinormal Mode (QNM) spectrum of black holes. We find that spectral instability may persist even after the potential well vanishes, giving rise to a distinct family of spectrally unstable QNMs that differ from the spectrally stable modes localized near the potential peak and associated with the photon sphere. Nevertheless, time-domain simulations reveal that early-time ringdown waveforms remain dominated by stable modes, while unstable modes have only a subdominant contribution. These results highlight the robustness of black hole spectroscopy, as the observable ringdown signal is primarily governed by the most stable QNMs.

gr-qc↗

arXiv2026

From Narrow to Panoramic Vision: Attention-Guided Cold-Start Reshapes Multimodal Reasoning

The cold-start initialization stage plays a pivotal role in training Multimodal Large Reasoning Models (MLRMs), yet its mechanisms remain insufficiently understood. To analyze this stage, we introduce the Visual Attention Score (VAS), an attention-based metric that quantifies how much a model attends to visual tokens. We find that reasoning performance is strongly correlated with VAS (r=0.9616): models with higher VAS achieve substantially stronger multimodal reasoning. Surprisingly, multimodal cold-start fails to elevate VAS, resulting in attention distributions close to the base model, whereas text-only cold-start leads to a clear increase. We term this counter-intuitive phenomenon Lazy Attention Localization. To validate its causal role, we design training-free interventions that directly modulate attention allocation during inference, performance gains of 1$-$2% without any retraining. Building on these insights, we further propose Attention-Guided Visual Anchoring and Reflection (AVAR), a comprehensive cold-start framework that integrates visual-anchored data synthesis, attention-guided objectives, and visual-anchored reward shaping. Applied to Qwen2.5-VL-7B, AVAR achieves an average gain of 7.0% across 7 multimodal reasoning benchmarks. Ablation studies further confirm that each component of AVAR contributes step-wise to the overall gains. The code, data, and models are available at https://github.com/lrlbbzl/Qwen-AVAR.

cs.CV↗

arXiv2026

EIMC: Efficient Instance-aware Multi-modal Collaborative Perception

Multi-modal collaborative perception calls for great attention to enhancing the safety of autonomous driving. However, current multi-modal approaches remain a ``local fusion to communication'' sequence, which fuses multi-modal data locally and needs high bandwidth to transmit an individual's feature data before collaborative fusion. EIMC innovatively proposes an early collaborative paradigm. It injects lightweight collaborative voxels, transmitted by neighbor agents, into the ego's local modality-fusion step, yielding compact yet informative 3D collaborative priors that tighten cross-modal alignment. Next, a heatmap-driven consensus protocol identifies exactly where cooperation is needed by computing per-pixel confidence heatmaps. Only the Top-K instance vectors located in these low-confidence, high-discrepancy regions are queried from peers, then fused via cross-attention for completion. Afterwards, we apply a refinement fusion that involves collecting the top-K most confident instances from each agent and enhancing their features using self-attention. The above instance-centric messaging reduces redundancy while guaranteeing that critical occluded objects are recovered. Evaluated on OPV2V and DAIR-V2X, EIMC attains 73.01\% AP@0.5 while reducing byte bandwidth usage by 87.98\% compared with the best published multi-modal collaborative detector. Code publicly released at https://github.com/sidiangongyuan/EIMC.

cs.CV↗

arXiv2026

Dense-Jump Flow Matching with Non-Uniform Time Scheduling for Robotic Policies: Mitigating Multi-Step Inference Degradation

Flow matching has emerged as a competitive framework for learning high-quality generative policies in robotics; however, we find that generalisation arises and saturates early along the flow trajectory, in accordance with recent findings in the literature. We further observe that increasing the number of Euler integration steps during inference counter-intuitively and universally degrades policy performance. We attribute this to (i) additional, uniformly spaced integration steps oversample the late-time region, thereby constraining actions towards the training trajectories and reducing generalisation; and (ii) the learned velocity field becoming non-Lipschitz as integration time approaches 1, causing instability. To address these issues, we propose a novel policy that utilises non-uniform time scheduling (e.g., U-shaped) during training, which emphasises both early and late temporal stages to regularise policy training, and a dense-jump integration schedule at inference, which uses a single-step integration to replace the multi-step integration beyond a jump point, to avoid unstable areas around 1. Essentially, our policy is an efficient one-step learner that still pushes forward performance through multi-step integration, yielding up to 23.7% performance gains over state-of-the-art baselines across diverse robotic tasks.

cs.RO↗