Searcharxiv⌕ Search

arXiv subjects

Yan Wang

Publications and source records attributed to Yan Wang.

At least 217 records · Page 12Linked to original sources
arXiv2026

ConsistRM: Improving Generative Reward Models via Consistency-Aware Self-Training

Generative reward models (GRMs) have emerged as a promising approach for aligning Large Language Models (LLMs) with human preferences by offering greater representational capacity and flexibility than traditional scalar reward models. However, GRMs face two major challenges: reliance on costly human-annotated data restricts scalability, and self-training approaches often suffer from instability and vulnerability to reward hacking. To address these issues, we propose ConsistRM, a self-training framework that enables effective and stable GRM training without human annotations. ConsistRM incorporates the Consistency-Aware Answer Reward, which produces reliable pseudo-labels with temporal consistency, thereby providing more stable model optimization. Moreover, the Consistency-Aware Critique Reward is introduced to assess semantic consistency across multiple critiques and allocates fine-grained and differentiated rewards. Experiments on five benchmark datasets across four base models demonstrate that ConsistRM outperforms vanilla Reinforcement Fine-Tuning (RFT) by an average of 1.5%. Further analysis shows that ConsistRM enhances output consistency and mitigates position bias caused by input order, highlighting the effectiveness of consistency-aware rewards in improving GRMs. Our implementation is available at https://github.com/yuliangCarmelo/ConsistRM.

cs.AI↗
arXiv2026

Training LLM Agents for Spontaneous, Reward-Free Self-Evolution via World Knowledge Exploration

Most agents today ``self-evolve'' by following rewards and rules defined by humans. However, this process remains fundamentally dependent on external supervision; without human guidance, the evolution stops. In this work, we train agents to possess an intrinsic meta-evolution capability to spontaneously learn about unseen environments prior to task execution. To instill this ability, we design an outcome-based reward mechanism that measures how much an agent's self-generated world knowledge improves its success rate on downstream tasks. This reward signal is used exclusively during the training phase to teach the model how to explore and summarize effectively. At inference time, the agent requires no external rewards or human instructions. It spontaneously performs native self-evolution to adapt to unknown environments using its internal parameters. When applied to Qwen3-30B and Seed-OSS-36B, this shift to native evolution yields a 20% performance increase on WebVoyager and WebWalker. Most strikingly, the generated world knowledge even enables a compact 14B Qwen3 model to outperform the unassisted Gemini-2.5-Flash, establishing a new paradigm for truly evolving agents.

cs.AI↗
arXiv2026

ReflectRM: Boosting Generative Reward Models via Self-Reflection within a Unified Judgment Framework

Reward Models (RMs) are critical components in the Reinforcement Learning from Human Feedback (RLHF) pipeline, directly determining the alignment quality of Large Language Models (LLMs). Recently, Generative Reward Models (GRMs) have emerged as a superior paradigm, offering higher interpretability and stronger generalization than traditional scalar RMs. However, existing methods for GRMs focus primarily on outcome-level supervision, neglecting analytical process quality, which constrains their potential. To address this, we propose ReflectRM, a novel GRM that leverages self-reflection to assess analytical quality and enhance preference modeling. ReflectRM is trained under a unified generative framework for joint modeling of response preference and analysis preference. During inference, we use its self-reflection capability to identify the most reliable analysis, from which the final preference prediction is derived. Experiments across four benchmarks show that ReflectRM consistently improves performance, achieving an average accuracy gain of +3.7 on Qwen3-4B. Further experiments confirm that response preference and analysis preference are mutually reinforcing. Notably, ReflectRM substantially mitigates positional bias, yielding +10.2 improvement compared with leading GRMs and establishing itself as a more stable evaluator. Our code is available at https://github.com/yuliangCarmelo/ReflectRM.

cs.AI↗
arXiv2026

Same Claim, Different Judgment: Benchmarking Scenario-Induced Bias in Multilingual Financial Misinformation Detection

Large language models (LLMs) have been widely applied across various domains of finance. Since their training data are largely derived from human-authored corpora, LLMs may inherit a range of human biases. Behavioral biases can lead to instability and uncertainty in decision-making, particularly when processing financial information. However, existing research on LLM bias has mainly focused on direct questioning or simplified, general-purpose settings, with limited consideration of the complex real-world financial environments and high-risk, context-sensitive, multilingual financial misinformation detection tasks MFMD. In this work, we propose MFMDScen, a comprehensive benchmark for evaluating behavioral biases of LLMs in MFMD across diverse economic scenarios. In collaboration with financial experts, we construct three types of complex financial scenarios: (i) role- and personality-based, (ii) role- and region-based, and (iii) role-based scenarios incorporating ethnicity and religious beliefs. We further develop a multilingual financial misinformation dataset covering English, Chinese, Greek, and Bengali. By integrating these scenarios with misinformation claims, MFMDScen enables a systematic evaluation of 22 mainstream LLMs. Our findings reveal that pronounced behavioral biases persist across both commercial and open-source models. This project is available at https://github.com/lzw108/FMD.

cs.CL↗
arXiv2026

Breaking the Training Barrier of Billion-Parameter Universal Machine Learning Interatomic Potentials

Universal Machine Learning Interatomic Potentials (uMLIPs), pre-trained on massively diverse datasets encompassing inorganic materials and organic molecules across the entire periodic table, serve as foundational models for quantum-accurate physical simulations. However, uMLIP training requires second-order derivatives, which lack corresponding parallel training frameworks; moreover, scaling to the billion-parameter regime causes explosive growth in computation and communication overhead, making its training a tremendous challenge. We introduce MatRIS-MoE, a billion-parameter Mixture-of-Experts model built upon invariant architecture, and {Janus}, a pioneering high-dimensional distributed training framework for uMLIPs with hardware-aware optimizations. Deployed across two Exascale supercomputers, our code attains a peak performance of 1.2/1.0 EFLOPS (24\%/{35.5\%} of theoretical peak) in single precision at over 90\% parallel efficiency, compressing the training of billion-parameter uMLIPs from weeks to hours. This work establishes a new high-water mark for AI-for-Science (AI4S) foundation models at Exascale and provides essential infrastructure for rapid scientific discovery.

cs.DC↗
arXiv2026

Intent Propagation Contrastive Collaborative Filtering

Disentanglement techniques used in collaborative filtering uncover interaction intents between nodes, improving the interpretability of node representations and enhancing recommendation performance. However, existing disentanglement methods still face two problems. First, they focus on local structural features derived from direct node interactions and overlook the comprehensive graph structure, which limits disentanglement accuracy. Second, the disentanglement process depends on backpropagation signals derived from recommendation tasks and lacks direct supervision, which may lead to biases and overfitting. To address these issues, we propose the Intent Propagation Contrastive Collaborative Filtering (IPCCF) algorithm. Specifically, we design a double helix message propagation framework to more effectively extract the deep semantic information of nodes, thereby improving the model's understanding of interactions between nodes. We also develop an intent message propagation method that incorporates graph structure information into the disentanglement process, thereby expanding the consideration scope of disentanglement. In addition, contrastive learning techniques are employed to align node representations derived from structure and intents, providing direct supervision for the disentanglement process, mitigating biases, and enhancing the model's robustness to overfitting. Experiments on three real data graphs illustrate the superiority of the proposed approach.

cs.IR↗
arXiv2026

AgentIAD: Agentic Industrial Anomaly Detection via Adaptive Memory Augmentation

Industrial anomaly detection (IAD) is challenging due to the subtle and highly localized nature of many defects, which single-pass vision--language models (VLMs) often fail to capture. Moreover, existing approaches lack mechanisms to actively acquire complementary evidence during inference. We propose AgentIAD, an agentic vision--language framework that enables iterative industrial inspection through a unified action space. The agent dynamically accesses two forms of memory during inspection: visual memory via the Perceptive Zoomer (PZ) for fine-grained local analysis, and retrieved memory via the Web Searcher (WS) and Comparative Retriever (CR) for external knowledge acquisition and cross-instance verification. This design allows the model to progressively gather evidence through multi-round perception--action reasoning. To effectively learn such policies under sparse supervision, AgentIAD adopts a two-stage training strategy: tool-aware supervised fine-tuning first initializes structured reasoning and memory-access behaviors, followed by agentic reinforcement learning to refine long-horizon decision policies. Extensive experiments show that, under the same backbone, AgentIAD improves classification accuracy by 5.92% over the previous state-of-the-art method on the MMAD benchmark while providing more reliable and interpretable anomaly analysis.

cs.CV↗
arXiv2026

Cognition-Inspired Dual-Stream Semantic Enhancement for Vision-Based Dynamic Emotion Modeling

The human brain constructs emotional percepts not by processing facial expressions in isolation, but through a dynamic, hierarchical integration of sensory input with semantic and contextual knowledge. However, existing vision-based dynamic emotion modeling approaches often neglect emotion perception and cognitive theories. To bridge this gap between machine and human emotion perception, we propose cognition-inspired Dual-stream Semantic Enhancement (DuSE). Our model instantiates a dual-stream cognitive architecture. The first stream, a Hierarchical Temporal Prompt Cluster (HTPC), operationalizes the cognitive priming effect. It simulates how linguistic cues pre-sensitize neural pathways, modulating the processing of incoming visual stimuli by aligning textual semantics with fine-grained temporal features of facial dynamics. The second stream, a Latent Semantic Emotion Aggregator (LSEA), computationally models the knowledge integration process, akin to the mechanism described by the Conceptual Act Theory. It aggregates sensory inputs and synthesizes them with learned conceptual knowledge, reflecting the role of the hippocampus and default mode network in constructing a coherent emotional experience. By explicitly modeling these neuro-cognitive mechanisms, DuSE provides a more neurally plausible and robust framework for dynamic facial expression recognition (DFER). Extensive experiments on challenging in-the-wild benchmarks validate our cognition-centric approach, demonstrating that emulating the brain's strategies for emotion processing yields state-of-the-art performance and enhances model interpretability.

cs.CV↗
arXiv2026

RPG-SAM: Reliability-Weighted Prototypes and Geometric Adaptive Threshold Selection for Training-Free One-Shot Polyp Segmentation

Training-free one-shot segmentation offers a scalable alternative to expert annotations where knowledge is often transferred from support images and foundation models. But existing methods often treat all pixels in support images and query response intensities models in a homogeneous way. They ignore the regional heterogeity in support images and response heterogeity in query.To resolve this, we propose RPG-SAM, a framework that systematically tackles these heterogeneity gaps. Specifically, to address regional heterogeneity, we introduce Reliability-Weighted Prototype Mining (RWPM) to prioritize high-fidelity support features while utilizing background anchors as contrastive references for noise suppression. To address response heterogeneity, we develop Geometric Adaptive Selection (GAS) to dynamically recalibrate binarization thresholds by evaluating the morphological consensus of candidates. Finally, an iterative refinement loop method is designed to polishes anatomical boundaries. By accounting for multi-layered information heterogeneity, RPG-SAM achieves a 5.56\% mIoU improvement on the Kvasir dataset. Code will be released.

cs.CV↗
arXiv2026

Screen, Cache, and Match: A Training-Free Causality-Consistent Reference Frame Framework for Human Animation

Human animation aims to generate temporally coherent and visually consistent videos over long sequences, yet modeling long-range dependencies while preserving frame quality remains challenging. Inspired by the human ability to leverage past observations for interpreting ongoing actions, we propose FrameCache, a training-free, causality-consistent reference frame framework. FrameCache explicitly converts historical generation results into causal guidance through two complementary mechanisms. First, at the reference level, a novel Screen-Cache-Match (SCM) strategy constructs a dynamic, high-quality reference memory, ensuring motion-consistent appearance guidance to reduce identity drift. Second, at the generative level, a Trajectory-Aware Autoregressive Generation (TAAG) mechanism aligns denoising trajectories across adjacent video chunks. This is achieved through an overlap-aware latent propagation and a dual-domain fusion strategy that seamlessly blends low-frequency structural layouts with high-frequency textural details. Extensive experiments on standard benchmarks demonstrate that FrameCache consistently improves temporal coherence and visual stability while integrating seamlessly with diverse diffusion baselines. Code will be made publicly available.

cs.GR↗
arXiv2026

Scalable continuous gravitational wave detection in PTA data with non-parametric red noise suppression and optimal pulsar selection

Bayesian methods for the detection of continuous gravitational waves (CGWs) in Pulsar Timing Array (PTA) data incur substantial computational costs that grow rapidly due to the number of noise and signal parameters characterizing the fitted model being proportional to the size of the PTA. This computational burden limits the scalability of these methods for large-scale PTAs comprising hundreds of pulsars anticipated from next-generation radio astronomy facilities. In this work, we introduce a computationally efficient frequentist method designed to circumvent this challenge. This is achieved by combining an adaptive spline fitting algorithm that non-parametrically suppresses red noise, thereby eliminating the need for complex noise modeling inherent to Bayesian methods, with a novel scheme for optimizing the subsets of pulsars included in the search. We quantify the performance of our method on a simulated dataset based on the NANOGrav 15-year data release and find that it achieves a performance comparable to that of Bayesian analysis: for a CGW signal with a signal-to-noise ratio of $\approx 10$, our method yields a relative characteristic strain error of 1.0\% and a frequency error of 0.072\% from the injected values by using the optimal pulsar selections, while the same errors are 1.7\% and 0.16\%, respectively, for the standard Bayesian analysis. At the same time, our analysis completes in less than 5 hours, in contrast to the 1-2 days required by Bayesian methods. This allows us to perform a rigorous study of our method using multiple data realizations and signal parameters, establishing it as an efficient and scalable tool for CGW searches with large-scale PTAs.

astro-ph.IM↗
arXiv2026

FinCriticalED: A Visual Benchmark for Financial Fact-Level OCR

Recent progress in multimodal large language models (MLLMs) has substantially improved document understanding, yet strong optical character recognition (OCR) performance on surface metrics does not guarantee faithful preservation of decision-critical evidence. This limitation is especially consequential in financial documents, where small visual errors can induce discrete shifts in meaning. To study this gap, we introduce FinCriticalED (Financial Critical Error Detection), a fact-centric visual benchmark for evaluating whether OCR and vision-language systems preserve financially critical evidence beyond lexical similarity. FinCriticalED contains 859 real-world financial document pages with 9,481 expert-annotated facts spanning five critical field types: numeric, temporal, monetary unit, reporting entity, and financial concept. We formulate the task as structured OCR with fact-level verification, and develop a Deterministic-Rule-Guided LLM-as-Judge protocol to assess whether model outputs preserve annotated facts in context. We benchmark 13 systems spanning OCR pipelines, specialized OCR VLMs, open-source MLLMs, and proprietary MLLMs. Results reveal a clear gap between lexical accuracy and factual reliability, with numerical values and monetary units emerging as the most vulnerable fact types, and critical errors concentrating in visually complex, mixed-layout documents with distinct failure patterns across model families. Overall, FinCriticalED provides a rigorous benchmark for trustworthy financial OCR and a practical testbed for evidence fidelity in high-stakes multimodal document understanding. Benchmark and dataset details available at https://the-finai.github.io/FinCriticalED/

cs.CV↗
arXiv2026

Fast-dVLA: Accelerating Discrete Diffusion VLA to Real-Time Performance

This paper proposes a novel approach to address the challenge that pretrained VLA models often fail to effectively improve performance and reduce adaptation costs during standard supervised finetuning (SFT). Some advanced finetuning methods with auxiliary training objectives can improve performance and reduce the number of convergence steps. However, they typically incur significant computational overhead due to the additional losses from auxiliary tasks. To simultaneously achieve the enhanced capabilities of auxiliary training with the simplicity of standard SFT, we decouple the two objectives of auxiliary task training within the parameter space, namely, enhancing general capabilities and fitting task-specific action distributions. To deliver this goal, we only need to train the model to converge on a small-scale task set using two distinct training strategies. The difference between the resulting model parameters can then be interpreted as capability vectors provided by auxiliary tasks. These vectors are then merged with pretrained parameters to form a capability-enhanced meta model. Moreover, when standard SFT is augmented with a lightweight orthogonal regularization loss, the merged model attains performance comparable to auxiliary finetuned baselines with reduced computational overhead. Experimental results demonstrate that this approach is highly effective across diverse robot tasks. Project page: https://chris1220313648.github.io/Fast-dVLA/

cs.RO↗
arXiv2026

Hackers or Hallucinators? A Comprehensive Analysis of LLM-Based Automated Penetration Testing

The rapid advancement of Large Language Models (LLMs) has created new opportunities for Automated Penetration Testing (AutoPT), spawning numerous frameworks aimed at achieving end-to-end autonomous attacks. However, despite the proliferation of related studies, existing research generally lacks systematic architectural analysis and large-scale empirical comparisons under a unified benchmark. Therefore, this paper presents the first Systematization of Knowledge (SoK) focusing on the architectural design and comprehensive empirical evaluation of current LLM-based AutoPT frameworks. At systematization level, we comprehensively review existing framework designs across six dimensions: agent architecture, agent plan, agent memory, agent execution, external knowledge, and benchmarks. At empirical level, we conduct large-scale experiments on 13 representative open-source AutoPT frameworks and 2 baseline frameworks utilizing a unified benchmark. The experiments consumed over 10 billion tokens in total and generated more than 1,500 execution logs, which were manually reviewed and analyzed over four months by a panel of more than 15 researchers with expertise in cybersecurity. By investigating the latest progress in this rapidly developing field, we provide researchers with a structured taxonomy to understand existing LLM-based AutoPT frameworks and a large-scale empirical benchmark, along with promising directions for future research.

cs.CR↗
arXiv2026

Ultrafast Non-Volatile Weyl LuminoMem for Mid-Infrared In-Memory Computing

Integrated optoelectronic systems strive to combine the logic/memory density of electronics with the bandwidth of photonics, but monolithic realization is impeded by the inefficient electronic-to-photonic interface. Current architectures rely on separate readout circuitry and modulators, creating bottlenecks in energy and latency, while existing direct transduction methods often compromise on switching speed or non-volatility. Here, we report an ultrafast, non-volatile optoelectronic memory, named LuminoMem, that integrates electrical storage and mid-infrared light emission in a single device. The device utilizes a floating-gate architecture, in which the Weyl semiconductor tellurium serves simultaneously as a charge-trapping storage layer and an emissive medium. This design enables nanosecond-scale electrical programming of non-volatile photoluminescence at 3.4 um, allowing direct optical access to stored states without external modulation. We demonstrate 4-bit (16-level) optical storage capacity and validate the device's performance through neural network simulations that achieve high accuracy on the Fashion-MNIST dataset. By effectively bridging the gap between electronic storage and mid-infrared photonics, the demonstrated mid-infrared LuminoMem provides a hardware foundation for promoting current computation efficiency and potential intelligent platforms that co-integrate computing, memory, and sensing capabilities.

cond-mat.mtrl-sci↗
arXiv2026

Training-Free Image Editing with Visual Context Integration and Concept Alignment

In image editing, it is essential to incorporate a context image to convey the user's precise requirements, such as subject appearance or image style. Existing training-based visual context-aware editing methods incur data collection effort and training cost. On the other hand, the training-free alternatives are typically established on diffusion inversion, which struggles with consistency and flexibility. In this work, we propose VicoEdit, a training-free and inversion-free method to inject the visual context into the pretrained text-prompted editing model. More specifically, VicoEdit directly transforms the source image into the target one based on the visual context, thereby eliminating the need for inversion that can lead to deviated trajectories. Moreover, we design a posterior sampling approach guided by concept alignment to enhance the editing consistency. Empirical results demonstrate that our training-free method achieves even better editing performance than the state-of-the-art training-based models.

cs.CV↗
arXiv2026

Rotatable Antenna-array-enhanced Direction-sensing for Low-altitude Communication Network: Method and Performance

In a practical multi-antenna receiver, each element of the receive antenna array has a directive antenna pattern, which is still not fully explored and investigated in academia and industry until now. When the emitter is deviated greatly from the normal direction of antenna element or is close to the null-point direction, the sensing energy by array will be seriously attenuated such that the direction-sensing performance is degraded significantly. To address such an issue, a rotatable array system is established with the directive antenna pattern of each element taken into account, where each element has the same antenna pattern. Then, the corresponding the Cramer-Rao lower bound (CRLB) is derived. Finally, a recursive rotation Root-MUSIC (RR-Root-MUSIC) direction-sensing method is proposed and its root-mean-squared-error (RMSE) performance is evaluated by the derived CRLB. Simulation results show that the proposed rotation method converges rapidly with about ten iterations, and make a significant enhancement on the direction-sensing accuracy in terms of RMSE when the target direction departs seriously far away from the normal vector of array. Compared with conventional Root-MUSIC, the sensing performance of the proposed RR-Root-MUSIC method is much closer to the CRLB.

eess.SP↗
arXiv2026

ViBA: Implicit Bundle Adjustment with Geometric and Temporal Consistency for Robust Visual Matching

Most existing image keypoint detection and description methods rely on datasets with accurate pose and depth annotations, limiting scalability and generalization, and often degrading navigation and localization performance. We propose ViBA, a sustainable learning framework that integrates geometric optimization with feature learning for continuous online training on unconstrained video streams. Embedded in a standard visual odometry pipeline, it consists of an implicitly differentiable geometric residual framework: (i) an initial tracking network for inter-frame correspondences, (ii) depth-based outlier filtering, and (iii) differentiable global bundle adjustment that jointly refines camera poses and feature positions by minimizing reprojection errors. By combining geometric consistency from BA with long-term temporal consistency across frames, ViBA enforces stable and accurate feature representations. We evaluate ViBA on EuRoC and UMA datasets. Compared with state-of-the-art methods such as SuperPoint+SuperGlue, ALIKED, and LightGlue, ViBA reduces mean absolute translation error (ATE) by 12-18% and absolute rotation error (ARE) by 5-10% across sequences, while maintaining real-time inference speeds (FPS 36-91). When evaluated on unseen sequences, it retains over 90% localization accuracy, demonstrating robust generalization. These results show that ViBA supports continuous online learning with geometric and temporal consistency, consistently improving navigation and localization in real-world scenarios.

cs.CV↗