Searcharxiv⌕ Search

arXiv subjects

Xiaoying Zhang

Publications and source records attributed to Xiaoying Zhang.

At least 19 recordsLinked to original sources
arXiv2026

Proximal Mediation Analysis with Unmeasured Treatment-Induced Confounding

Mediation analysis provides a central framework for elucidating causal mechanisms, yet its application is often impeded by treatment-induced confounding, under which the widely used natural mediation effects are generally unidentifiable. Interventional effects have been proposed as an alternative when these confounders are observable; however, identifying and estimating interventional effects remains challenging when confounders are unmeasured. In this paper, we address this issue by using observed variables as proxies for unmeasured treatment-induced confounders. We establish four proximal identification results and develop a multiply robust, semiparametric locally efficient estimator that accommodates flexible machine learning methods for nuisance parameter estimation. The proposed approach is illustrated through simulation studies and a real-data application evaluating racial disparities in life satisfaction mediated by discrimination.

stat.ME↗
arXiv2026

RetroAgent: From Solving to Evolving via Retrospective Dual Intrinsic Feedback

Standard reinforcement learning (RL) for large language model (LLM) agents primarily optimizes extrinsic task rewards, often favoring isolated task completion over continual adaptation. This paradigm can cause premature convergence to suboptimal policies and leaves useful experience only implicitly encoded in model parameters, limiting its retrieval and reuse for future decisions. We introduce RetroAgent, an online RL framework that trains agents to master interactive environments not merely by solving tasks, but by evolving across episodes. Inspired by human retrospective self-improvement, RetroAgent augments extrinsic rewards with hindsight-generated dual intrinsic feedback: (1) Intrinsic Numerical Feedback, which rewards beneficial exploration by measuring incremental subtask progress relative to prior attempts; and (2) Intrinsic Language Feedback which distills successes and failures into reusable textual lessons for explicit experience reuse. To leverage these lessons effectively, we propose Similarity & Utility-Aware Upper Confidence Bound (SimUtil-UCB), a retrieval strategy that balances semantic relevance, historical utility, and exploration. Across four challenging agentic benchmarks, RetroAgent achieves new state-of-the-art performance, outperforming GRPO by +18.3% on ALFWorld, +15.4% on WebShop, +27.1% on Sokoban, and +8.9% on MineSweeper, while demonstrating strong test-time adaptation and out-of-distribution generalization.

cs.AI↗
arXiv2026

Critique-GRPO: Advancing LLM Reasoning with Natural Language and Numerical Feedback

Recent advances in reinforcement learning (RL) using numerical rewards have significantly enhanced the complex reasoning capabilities of large language models (LLMs). However, we identify three fundamental limitations of purely numerical feedback: performance plateaus, ineffective spontaneous self-reflection, and persistent failures. We show that plateaued RL models can successfully refine failed solutions when given natural language critiques. Motivated by this, we propose Critique-GRPO, an online RL framework that integrates both natural language and numerical feedback for policy optimization. This approach enables LLMs to learn simultaneously from initial responses and critique-guided refinements, effectively internalizing the exploration benefits of both stages. Extensive experiments show that Critique-GRPO outperforms all compared supervised and RL-based fine-tuning methods, achieving average Pass@1 improvements of approximately +15.0-21.6% on various Qwen models and +7.3% on Llama-3.2-3B-Instruct across eight challenging reasoning tasks. Notably, Critique-GRPO facilitates effective self-improvement through self-critiquing, achieving substantial gains over GRPO, e.g., a +16.7% Pass@1 improvement on AIME 2024. The code and models are released at: https://github.com/zhangxy-2019/critique-GRPO

cs.CL↗
arXiv2026

Leveraging Error Diversity in Group Rollouts for Reinforcement Learning

Reinforcement Learning from Verifiable Rewards (RLVR) typically samples multiple responses per prompt and assigns binary rewards based on individual correctness, yet the collective structure of the group output, specifically the distribution of errors, is largely discarded. We identify this as a missed opportunity: empirical analysis reveals that error diversity within a group is a strong predictor of training success, with problems eliciting diverse wrong answers benefiting substantially more from RLVR than those producing homogeneous failures. Motivated by this observation, we propose Error Diversity Advantage Shaping (EDAS), a lightweight, algorithm-agnostic technique that modulates the advantage signal for incorrect rollouts based on intra-group error diversity. EDAS amplifies penalties for dominant, repeated errors and attenuates penalties for rare, exploratory ones, thereby encouraging the model to maintain diverse reasoning paths and discouraging error perseveration. Crucially, EDAS operates as a simple post-hoc adjustment that can be seamlessly integrated into any RLVR algorithm. We validate EDAS on top of several mainstream RLVR methods across a series of models and seven challenging math benchmarks, demonstrating consistent improvements. Notably, EDAS yields an average improvement of 6.29 points over DAPO on Qwen3-8B across seven benchmarks, confirming that exploiting the latent information in group rollouts is a broadly effective strategy for strengthening RLVR.

cs.LG↗
arXiv2026

Right Makes Might: Aligning Verified Hidden States Empowers RL Reasoning

Reinforcement Learning from Verifiable Rewards (RLVR) has become the dominant approach for improving mathematical reasoning in large language models, yet current methods reduce each correct rollout to a single reward bit, ignoring the geometric structure shared among their hidden states. Investigating this structure, we find that at the anchor token (the position immediately before the answer marker), correct rollouts converge naturally because they must produce the same answer (cosine similarity ~0.84), yet each retains residual variance from its unique reasoning path. Encouraging full alignment at this point pushes the model to extract a unified "correct decision" representation, reducing sensitivity to which reasoning path was taken. Based on this observation, we propose Hidden-Align, an auxiliary loss function that aligns the last-layer hidden states of correct rollouts at the anchor token during RL training, with zero overhead in both training and inference. On eight mathematical reasoning benchmarks, Hidden-Align improves average pass@1 over the DAPO baseline by 3.8, 6.2, and 5.4 percentage points on Qwen3-1.7B, 4B, and 14B respectively, with consistent pass@k gains across all three scales, supported by ablations on loss type, anchor position, layer depth, and loss weight.

cs.LG↗
arXiv2026

Step-wise Rubric Rewards for LLM Reasoning

Reinforcement Learning with Verifiable Rewards (RLVR) is widely used to improve reasoning in large language models, but rewards only final-answer correctness with no supervision over intermediate steps. Rubric-based methods such as Rubrics as Rewards (RaR) introduce finer-grained supervision by scoring rollouts against structured criteria, yet the rubric scores are still aggregated into a single scalar applied to the entire response, causing three weaknesses: loss of multi-criterion structure, uniform supervision of correct and incorrect steps, and reward hacking through unbounded self-correction. On 1,000 problems, we find 18.2% of steps in correct-answer responses are wrong yet positively rewarded, while 49.9% of steps in incorrect-answer responses are correct yet penalized. We introduce Step-wise Rubrics as Rewards (SRaR), an RLVR framework that (i) uses an LLM judge to attribute each rubric item to a specific reasoning step, (ii) normalizes per-step rubric scores across rollouts so only steps whose quality varies produce a learning signal, and (iii) combines the per-step reward with the outcome reward through a decoupled advantage estimator that keeps the outcome baseline stable. We further build a 16K-problem rubric dataset by contrastively distilling rubric items from correct and flawed reasoning paths sampled from a strong model. Across six mathematical reasoning benchmarks, SRaR improves average accuracy over RaR by 3.57 points on Qwen3-8B and 2.75 points on Qwen3-32B, raises the Faithful Reasoning Rate on AIME 2025 from 34.5% to 46.7%, and reduces self-correction looping from 48.1% to 26.5%.

cs.LG↗
arXiv2026

A Geometry Map-Based Site-Specific Propagation Channel Model for Urban Scenarios

With the rapid deployments of 5G and 6G networks, accurate modeling of urban radio propagation has become critical for system design and network planning. However, conventional statistical or empirical models fail to fully capture the influence of detailed geometric features on site-specific channel variances in dense urban environments. In this paper, we propose a geometry map-based propagation channel model that directly extracts key parameters from a 3D geometry map and incorporates the Uniform Theory of Diffraction (UTD) to recursively compute multiple diffraction fields, thereby enabling accurate prediction of site-specific large-scale path loss and time-varying Doppler characteristics in urban scenarios. A well-designed identification algorithm is developed to efficiently detect buildings that significantly affect signal propagation. The proposed model is validated using urban measurement data, showing excellent agreement of path loss in both line-of-sight (LOS) and nonline-of-sight (NLOS) conditions. In particular, for NLOS scenarios with complex diffractions, it outperforms the 3GPP and simplified models, reducing the RMSE by 7.1 dB and 3.18 dB, respectively. Doppler analysis further demonstrates its accuracy in capturing time-varying propagation characteristics, confirming the scalability and generalization of the model in urban environments.

eess.SP↗
arXiv2026

A Novel Site-Specific Inference Model for Urban Canyon Channels: From Measurements to Modeling

With the rapid development of intelligent transportation and smart city applications, urban canyon has become a critical scenario for the design and evaluation of wireless communication systems. Due to its unique environmental layout, the channel characteristics in urban canyon are strongly a street geometry and building distribution, thereby exhibiting significant site-specific channel condition. However, this feature has not been well captured in existing channel models. In this paper, we propose a site-specific channel inference model based on environmental geometry, the model is parameterized using sub-6GHz channel measurements. Multipath components (MPCs) are extracted and clustered according to geometric propagation, which are explicitly derived from the influence of canyon width, thereby establishing an interpretable mapping between the physical environment and statistical characteristics of MPCs. A step-by-step implementation scheme is presented. Subsequently, the proposed site-specific channel inference model is validated by comparing second-order statistics of channels, derived from the model and measurements. The results show that the proposed model achieves high accuracy and robustness in different urban canyon scenarios.

eess.SP↗
arXiv2026

Discovery of unobservable parameters via physical embedding

Recovering a source signal from indirect measurements often requires estimating latent parameters, such as wireless channel states or MRI coil sensitivities, that cannot be directly observed. Here, we introduce Physics-Embedded Inverse Learning (PEIL), in which a learned estimator predicts these parameters and a fixed, physics-based inverse operator uses them to reconstruct the signal, so that training requires only the source signal as supervision. In systems where multiple parameter combinations can reconstruct the signal equally well, the estimator exploits this freedom to coordinate parameters that compensate for residual modelling errors rather than match ground-truth parameters. In high-mobility wireless communications, PEIL discovers task-optimal configurations that outperform baselines given access to ground-truth parameters, enabling zero-shot generalisation and over 20-fold reduction in training data relative to supervised baselines. To test whether these properties extend across physical domains, we apply PEIL to parallel MRI, where it discovers physically interpretable coil sensitivity maps without calibration scans, yielding reconstructions grounded purely in acquired measurements. These results demonstrate that non-identifiability, conventionally a liability, becomes a resource when the learning objective targets reconstruction quality rather than parameter accuracy.

eess.SP↗
arXiv2026

Computing Maximal Per-Record Leakage and Leakage-Distortion Functions for Privacy Mechanisms under Entropy-Constrained Adversaries

The exponential growth of data collection necessitates robust privacy protections that preserve data utility. We address information disclosure against adversaries with bounded prior knowledge, modeled by an entropy constraint $H(X) \geq b$. Within this information privacy framework -- which replaces differential privacy's independence assumption with a bounded-knowledge model -- we study three core problems: maximal per-record leakage, the primal leakage-distortion tradeoff (minimizing worst-case leakage under distortion $D$), and the dual distortion minimization (minimizing distortion under leakage constraint $L$). These problems resemble classical information-theoretic ones (channel capacity, rate-distortion) but are more complex due to high dimensionality and the entropy constraint. We develop efficient alternating optimization algorithms that exploit convexity-concavity duality, with theoretical guarantees including local convergence for the primal problem and convergence to a stationary point for the dual. Experiments on binary symmetric channels and modular sum queries validate the algorithms, showing improved privacy-utility tradeoffs over classical differential privacy mechanisms. This work provides a computational framework for auditing privacy risks and designing certified mechanisms under realistic adversary assumptions.

cs.CR↗
arXiv2026

Integrated Channel Sounding and Communication: Requirements, Architecture, Challenges, and Key Technologies

Channel models are essential for the design, evaluation, and optimization of wireless communication systems. The emerging space-air-ground-sea integrated network (SAGSIN), characterized by diverse service applications and extended-spectrum operations, places even greater demands on highly accurate channel models. However, conventional channel sounding is limited by generalized measurement campaigns, inadequate cross-band consistency, and insufficient real-time adaptability, making it unable to meet the needs of SAGSIN for scenario-specific and high-precision channel modeling. To address this challenge, we propose a novel technological framework, termed integrated channel sounding and communication (ICSC). By deeply integrating sounding and communication, the ICSC enables efficient and real-time acquisition of dynamic channel characteristics during communication processes, supporting fine-grained site- and scenario-specific measurements. Furthermore, leveraging artificial intelligence techniques, ICSC can identify channel conditions and adapt waveform parameters in real-time according to scenario variations, which in turn enhances communication performance. This article first introduces the fundamental principles of the ICSC framework, elaborates on its core concepts and key advantages, and demonstrates its feasibility through the development of an integrated verification system (IVS). Subsequently, the potential applications and opportunities of the ICSC are analyzed in depth, followed by a discussion of its future development directions and remaining challenges.

eess.SP↗
arXiv2026

Visual Generation Unlocks Human-Like Reasoning through Multimodal World Models

Humans construct internal world models and reason by manipulating the concepts within these models. Recent advances in AI, particularly chain-of-thought (CoT) reasoning, approximate such human cognitive abilities, where world models are believed to be embedded within large language models. Expert-level performance in formal and abstract domains such as mathematics and programming has been achieved in current systems by relying predominantly on verbal reasoning. However, they still lag far behind humans in domains like physical and spatial intelligence, which require richer representations and prior knowledge. The emergence of unified multimodal models (UMMs) capable of both verbal and visual generation has therefore sparked interest in more human-like reasoning grounded in complementary multimodal pathways, though their benefits remain unclear. From a world-model perspective, this paper presents the first principled study of when and how visual generation benefits reasoning. Our key position is the visual superiority hypothesis: for certain tasks--particularly those grounded in the physical world--visual generation more naturally serves as world models, whereas purely verbal world models encounter bottlenecks arising from representational limitations or insufficient prior knowledge. Theoretically, we formalize internal world modeling as a core component of CoT reasoning and analyze distinctions among different forms of world models. Empirically, we identify tasks that necessitate interleaved visual-verbal CoT reasoning, constructing a new evaluation suite, VisWorld-Eval. Controlled experiments on a state-of-the-art UMM show that interleaved CoT significantly outperforms purely verbal CoT on tasks that favor visual world modeling, but offers no clear advantage otherwise. Together, this work clarifies the potential of multimodal world modeling for more powerful, human-like multimodal AI.

cs.AI↗
arXiv2025

Video-R1: Reinforcing Video Reasoning in MLLMs

Inspired by DeepSeek-R1's success in eliciting reasoning abilities through rule-based reinforcement learning (RL), we introduce Video-R1 as the first attempt to systematically explore the R1 paradigm for incentivizing video reasoning within multimodal large language models (MLLMs). However, directly applying RL training with the GRPO algorithm to video reasoning presents two primary challenges: (i) a lack of temporal modeling for video reasoning, and (ii) the scarcity of high-quality video-reasoning data. To address these issues, we first propose the T-GRPO algorithm, which encourages models to utilize temporal information in videos for reasoning. Additionally, instead of relying solely on video data, we incorporate high-quality image-reasoning data into the training process. We have constructed two datasets: Video-R1-CoT-165k for SFT cold start and Video-R1-260k for RL training, both comprising image and video data. Experimental results demonstrate that Video-R1 achieves significant improvements on video reasoning benchmarks such as VideoMMMU and VSI-Bench, as well as on general video benchmarks including MVBench and TempCompass, etc. Notably, Video-R1-7B attains a 37.1% accuracy on video spatial reasoning benchmark VSI-bench, surpassing the commercial proprietary model GPT-4o. All code, models, and data are released in: https://github.com/tulerfeng/Video-R1.

cs.CV↗
arXiv2025

Generative Universal Verifier as Multimodal Meta-Reasoner

We introduce Generative Universal Verifier, a novel concept and plugin designed for next-generation multimodal reasoning in vision-language models and unified multimodal models, providing the fundamental capability of reflection and refinement on visual outcomes during the reasoning and generation process. This work makes three main contributions: (1) We build ViVerBench, a comprehensive benchmark spanning 16 categories of critical tasks for evaluating visual outcomes in multimodal reasoning. Results show that existing VLMs consistently underperform across these tasks, underscoring a substantial gap from human-level capability in reliable visual verification. (2) We design two automated pipelines to construct large-scale visual verification data and train OmniVerifier-7B, the first omni-capable generative verifier trained for universal visual verification and achieves notable gains on ViVerBench(+8.3). Through training, we identify three atomic capabilities in visual verification and demonstrate how they generalize and interact synergistically. (3) We propose OmniVerifier-TTS, a sequential test-time scaling paradigm that leverages the universal verifier to bridge image generation and editing within unified models, enhancing the upper bound of generative ability through iterative fine-grained optimization. Beyond generation, we extend universal verifier to broader world-modeling interleaved reasoning scenarios. Empirically, OmniVerifier-TTS achieves improvements on T2I-ReasonBench(+3.7), and GenEval++(+4.3), outperforming existing parallel test-time scaling methods, such as Best-of-N. By endowing multimodal reasoning with reliable visual verification, OmniVerifier advances both reliable reflection during generation and scalable test-time refinement, marking a step toward more trustworthy and controllable next-generation reasoning systems.

cs.CV↗
arXiv2025

Affine Frequency Division Multiplexing for Communication and Channel Sounding: Requirements, Challenges, and Key Technologies

Channel models are crucial for theoretical analysis, performance evaluation, and deployment of wireless communication systems. Traditional channel sounding systems are insufficient for handling the dynamic changes of channels in the next-generation space-air-ground-sea integrated networks (SAGSIN), which often results in outdated channel models that fail to provide reliable prior information for communication systems. To address this challenge, this paper proposes an integrated channel sounding and communication (ICSC) method as a practical solution. Unlike orthogonal frequency division multiplexing, affine frequency division multiplexing (AFDM) provides a full delay-Doppler representation of the channel, achieving optimal diversity in time-frequency doubly dispersive channels and effectively addressing the aforementioned challenges. Thus, we investigate the fundamental principles of AFDM, showing how it enables simultaneous communication and channel sounding, and explore key performance metrics for both functionalities. We also clarify the distinction and relationship between channel sounding, estimation, tracking and scatterer sensing. Additionally, several potential application scenarios for AFDM-ICSC are explored. Finally, we highlight the key challenges in implementing AFDM-ICSC, outline future research directions, and provide valuable insights for the continued development of this technology.

eess.SP↗
arXiv2025

Discrepancies in Pedestrian Crossing of Static vs. Dynamic Crowds: An Experimental Study

In this paper, we investigate disparities in pedestrian crossing behaviors within static and dynamic crowds through experimental analysis. First, qualitative trajectory observations revealed significant behavioral differences in static and dynamic contexts. To quantitatively assess these discrepancies, we introduced a density metric termed the swarm factor. In static contexts, limited variations in speed and swarm factor were observed, which may contribute to the formation of cross-channels, a phenomenon of pedestrian self-organization (tactical level). In contrast, speed and swarm factor exhibited inverse synchronization in dynamic contexts, indicating density-dependent behavioral adaptation (operational level). Finally, orthogonal velocity analysis demonstrated a fundamental pattern in crossing motions: as global density increased, both instantaneous velocity and crossing velocity decreased, while transverse velocity remained stable.

physics.soc-ph↗
arXiv2025

Will Pre-Training Ever End? A First Step Toward Next-Generation Foundation MLLMs via Self-Improving Systematic Cognition

Recent progress in (multimodal) large language models ((M)LLMs) has shifted focus from pre-training to inference-time computation and post-training optimization, largely due to concerns over the availability of high-quality human data. However, these strategies alone are insufficient to drive substantial model improvements. We argue that effective model advancement requires strong synergy among pre-training, inference-time computation, and post-training optimization. In this paper, we introduce Self-Improving cognition (SIcog), a self-learning framework for constructing next-generation foundation MLLMs by imparting multimodal knowledge and enhancing systematic cognitive capabilities through multimodal pre-training with self-generated data. Specifically, we propose Chain-of-Description for step-by-step visual understanding and integrate structured Chain-of-Thought (CoT) reasoning to support in-depth multimodal reasoning. SIcog first equips a base model with systematic perception and reasoning using minimal external supervision. The enhanced models then generate candidate image captions and CoT reasoning responses for unlabeled images and image-question pairs across diverse tasks, which are filtered through a semantic-similarity-guided self-consistency mechanism. These high-quality, self-generated samples enable large-scale multimodal pre-training, creating a self-improvement loop. Experiments demonstrate SIcog's effectiveness in developing MLLMs with enhanced multimodal cognition. Using only 213K self-generated pre-training samples, SIcog achieves significant improvements, including +3.6% on MMStar and +3.5% on AI2D, outperforming previous pre-training approaches. When combined with post-training techniques for CoT reasoning, SIcog yields +9% gains on MMVet and +8.5% on ScienceQA.

cs.CV↗