Searcharxiv⌕ Search

arXiv subjects

Chen Zhao

Publications and source records attributed to Chen Zhao.

At least 73 records · Page 4Linked to original sources
arXiv2026

Reading, Not Thinking: Understanding and Bridging the Modality Gap When Text Becomes Pixels in Multimodal LLMs

Multimodal large language models (MLLMs) can process text presented as images, yet they often perform worse than when the same content is provided as textual tokens. We systematically diagnose this "modality gap" by evaluating seven MLLMs across seven benchmarks in five input modes, spanning both synthetically rendered text and realistic document images from arXiv PDFs to Wikipedia pages. We find that the gap is highly sensitive to rendering choices such as font and resolution, and that natural document images often exhibit much smaller gaps, suggesting the performance difference partly reflects evaluation artifacts rather than fundamental limitations. Through a grounded-theory error analysis of over 4,000 examples, we identify the primary cause: image input alone suppresses reasoning effort, with models producing 5--19x shorter outputs that skip step-by-step computation or reasoning. The reluctance to reason, not a failure of perception or knowledge retrieval, drives the performance gap, particularly on tasks requiring multi-step reasoning. We show that a simple, lightweight on-policy self-distillation method by fine-tuning models on their own text-mode reasoning traces paired with image inputs closes this gap, raising image-mode accuracy to match or exceed text-mode performance with over 50\% improvement, and the gains transfer to unseen benchmarks without catastrophic forgetting. Overall, our results and analyses provide a systematic understanding of the modality gap and suggest a practical path toward improving visual text understanding in multimodal language models.

cs.CL↗
arXiv2026

DXA-Derived Skeletal Phenotypes and Hip Fracture Risk: A Backdoor-Adjusted Causal Analysis

Purpose: To compare dual-energy X-ray absorptiometry (DXA)-derived hip skeletal phenotypes in relation to hip fracture risk using prespecified confounder adjustment and to assess whether phenotypes ranked by their backdoor-adjusted average treatment effects (ATEs) improve risk stratification. Methods: We analyzed 21,098 UK Biobank participants with linked health records, hip DXA-derived skeletal measures, and prespecified covariates. Sixteen phenotypes spanning bone mineral content (BMC), bone mineral density (BMD), and T-score across hip-related regions were evaluated. Confounder selection was guided by a prespecified directed acyclic graph (DAG). Backdoor-adjusted ATEs were estimated on the absolute risk-difference scale per standard deviation (SD) increase. Effect heterogeneity was evaluated for total femur BMD, and downstream prediction was assessed using clinical variables combined with phenotypes ranked by ATE magnitude. Results: Among 21,098 participants, 115 had hip fractures. All 16 phenotypes showed negative backdoor-adjusted ATEs per SD increase. The largest ATEs were observed for total femur BMC and total femur BMD, each with a risk difference of -0.0047, corresponding to approximately 4.7 fewer hip fractures per 1,000 participants per SD higher phenotype value. Conditional effects of total femur BMD were stronger among older participants and those with lower BMI. In prediction, clinical variables plus the top 11 ATE-ranked phenotypes achieved higher AUC than FRAX with femoral neck BMD (0.842 vs. 0.709), with higher sensitivity (0.748 vs. 0.443) and similar specificity (0.793 vs. 0.777). Conclusion: DXA-derived hip skeletal phenotypes differed in their backdoor-adjusted ATEs. Phenotype-level causal evaluation may help identify informative DXA measures for risk stratification.

q-bio.QM↗
arXiv2026

VisualLeakBench: Reproducible Action-Boundary Propagation Failures in Vision-Language Agents

Vision-language agents increasingly consume screenshots, documents, and user interfaces before writing to memory, sending messages, or invoking external tools. We study a concrete failure mode in this setting: action-boundary propagation, where sensitive or unsafe visible text is copied from an image into downstream tool arguments. We present VisualLeakBench, a diversified 500-image benchmark spanning UI, chat, document, form, and dashboard scenes, and evaluate a stratified 100-image agent subset with four production VLM systems under two workflows: note capture and external handoff. At baseline, target strings are propagated into tool arguments in 78.8% of PII cases and 85.5% of rendered unsafe-text cases. Under a defensive system prompt, rendered unsafe-text propagation remains high at 52.6%, while PII tool propagation falls to 2.0%, largely by suppressing tool use rather than preserving utility. Rates are tool-surface dependent: search-like tools suppress PII propagation, but rendered unsafe text still crosses tool boundaries. We measure visual-to-tool propagation rather than downstream instruction execution. We additionally provide a labeled-target oracle upper-bound diagnostic that localizes most failures at the tool boundary while leaving response-side leakage as residual risk.

cs.CV↗
arXiv2026

Transversal architecture for megaquop-scale quantum simulation with neutral atoms

Quantum computing experiments have made remarkable progress in demonstrating key components of quantum error correction, a prerequisite for scalable quantum computation. While we anticipate the arrival of early fault-tolerant quantum hardware capable of a million reliable quantum operations, the cost of preparing low-noise `magic resource states' presents a formidable challenge. The recently proposed partially-fault-tolerant architecture based on a space-time efficient analog rotation (STAR) approach attempts to address this challenge by using post-selection to prepare low-noise, small-angle magic states. Its proposed physical implementation, however, assumes fixed qubit connectivity, resulting in implementation costs closer to leading fully-fault-tolerant approaches. Here, we propose the transversal STAR architecture and co-design it with neutral-atom quantum hardware, deriving significant savings in logical layout, time, and space overhead. Through circuit-level simulations, we derive the logical noise model for surface-code-based transversal STAR gadgets and verify their composability. At its limit, the transversal STAR architecture can efficiently simulate local Hamiltonians with a total simulation volume exceeding 600. Achieving this limit would require approximately 10,000 physical qubits at a physical error rate of $10^{-3}$. This is equivalent to a fully-fault-tolerant computation requiring over $10^6$-$10^7$ $T$ gates. Finally, we extend the transversal STAR architecture to high-rate quantum codes, demonstrating how a limited set of highly parallel transversal Clifford gates and generalized small-angle magic injection can be utilized for effective quantum simulation. We anticipate that the co-designed transversal STAR architecture could substantially reduce the physical resources necessary for early-fault-tolerant quantum simulation at the megaquop scale.

quant-ph↗
arXiv2026

LUVE : Latent-Cascaded Ultra-High-Resolution Video Generation with Dual Frequency Experts

Recent advances in video diffusion models have significantly improved visual quality, yet ultra-high-resolution (UHR) video generation remains a formidable challenge due to the compounded difficulties of motion modeling, semantic planning, and detail synthesis. To address these limitations, we propose \textbf{LUVE}, a \textbf{L}atent-cascaded \textbf{U}HR \textbf{V}ideo generation framework built upon dual frequency \textbf{E}xperts. LUVE employs a three-stage architecture comprising low-resolution motion generation for motion-consistent latent synthesis, video latent upsampling that performs resolution upsampling directly in the latent space to mitigate memory and computational overhead, and high-resolution content refinement that integrates low-frequency and high-frequency experts to jointly enhance semantic coherence and fine-grained detail generation. Extensive experiments demonstrate that our LUVE achieves superior photorealism and content fidelity in UHR video generation, and comprehensive ablation studies further validate the effectiveness of each component. The project is available at \href{https://unicornanrocinu.github.io/LUVE_web/}{https://github.io/LUVE/}.

cs.CV↗
arXiv2026

Coupled Hierarchical Search over Topology and Execution for Agentic Workflow Synthesis

Although structured workflows empower Large Language Models (LLMs) to tackle complex problems, automating their creation is severely hindered by a vast combinatorial search space, frequently resulting in inflexible and resource-heavy offline training dependencies. To address this, we conceptualize workflow generation as an intertwined topology-and-execution search paradigm, where the broader topological layer dictates subtask boundaries and lower-level execution outcomes actively reshape the topology itself. Building on this foundation, we introduce HierFlow, a training-free, test-time hierarchical search architecture that automates agentic workflow design by merging feedback-guided topology adjustments with a fast, MCTS-inspired tree search for sub-workflow optimization. HierFlow maximizes efficiency through an intelligent gating module that selectively triggers execution-level searches based on contextual necessity, a mechanism we further support with an in-depth analysis detailing how varying degrees of cross-task coupling impact the effectiveness of hierarchical splitting. Comprehensive testing across question answering, mathematical reasoning, and code generation benchmarks confirms that HierFlow consistently outperforms strong baselines, delivering an optimal balance of high-quality results and computational efficiency without any additional training overhead.

cs.AI↗
arXiv2026

MIITA: Memory-Induced Inference-Time Adaptation for Continual Learning with Small Language Models

Continual learning (CL) is essential for small language models (SLMs) to adapt to evolving real-world needs in resource-constrained deployments. However, directly updating their limited parameter space causes catastrophic forgetting. While memory-based methods naturally address this by decoupling knowledge retention from parameters, existing approaches designed for large language models (LLMs) rely on abundant storage and strong in-context reasoning that SLMs lack. To address these challenges, we propose MIITA, a Memory-Induced Inference-Time Adaptation framework for supervised CL under constrained storage. MIITA stores supervised experiences as compact correction-direction prototypes with semantic anchors, and retrieves them at inference time using semantic and uncertainty-based cues. The retrieved directions are applied through gated temporary hidden-state adaptation, enabling non-destructive reuse of past supervision without backbone updates, prompt extensions, or test-time backpropagation. A local theoretical analysis links this design to first-order loss reduction, uncertainty-guided retrieval, and directional coverage for retaining old-stage knowledge. Extensive experiments across diverse supervised CL settings show that MIITA consistently improves final performance and mitigates forgetting under fixed memory budgets.

cs.AI↗
arXiv2026

Uncertainty-Guided Conservative Propagation for Structured Inference in Vessel Segmentation

Accurate vessel segmentation is essential for medical image analysis, yet remains challenging due to complex vascular patterns and imaging ambiguity. Most deep models rely on single-pass prediction, limiting their ability to refine uncertain or disconnected regions during inference. To address this limitation, we propose Uncertainty-Guided Conservative Propagation (UGCP), a general plug-in module for vessel segmentation. Instead of directly using a one-shot output as the final prediction, UGCP performs a small number of logit-space update steps to refine the segmentation through local predictions interaction. Predictive uncertainty guides reliable regions to support ambiguous regions, while structure-aware modulation and source-based stabilization reduce unreliable propagation and excessive drift. The module is differentiable and can be trained end-to-end with different segmentation networks. We evaluate UGCP on four public vessel segmentation datasets covering 2D and 3D tasks, including retinal vessel, coronary artery, and cerebral vessel segmentation. Experiments with convolutional neural network-based and Transformer-based backbones show consistent improvements in Dice similarity coefficient, centerline Dice, and 95th percentile Hausdorff distance. Further analysis demonstrates that UGCP reduces vessel disconnections and improves structural consistency with limited additional computation. The code will be made available at https://github.com/chenzhao2023/UGC_PR.

cs.CV↗
arXiv2026

The Unlearnability Phenomenon in RLVR for Language Models

Reinforcement Learning with Verifiable Reward (RLVR) has proven effective in improving Large Language Model's (LLM) reasoning ability. However, the learning dynamics of RLVR remain underexplored. In this paper, we reveal a counterintuitive phenomenon: among hard examples that the model initially struggles with, a substantial subset remains unlearnable even when correct rollouts are present. To understand the phenomenon, we first demonstrate that existing optimization and sampling techniques fail to resolve unlearnability. With cross-example gradient analysis, we show that unlearnable examples have fundamental representation issue, characterized by low gradient similarity with the rest of the examples and ungeneralizable reasoning patterns. We further show that representation flaws are difficult to mitigate in RL, as data augmentation does not improve gradient similarity. Our study provides the first systematic characterization of unlearnable data in RLVR training and reveals fundamental limitations in current RL approaches for reasoning tasks. Code and data are available at \url{https://github.com/yulinchen99/unlearnability-rlvr}.

cs.LG↗
arXiv2026

GraphFlash: Enabling Fast and Elastic Graph Processing on Serverless Infrastructure

Graph processing systems are essential for analyzing large-scale data with complex relationships, yet most existing frameworks rely on statically provisioned clusters, resulting in poor elasticity and inefficient resource utilization under dynamic workloads. Serverless computing offers automatic scaling and fine-grained billing, but existing serverless graph systems suffer from performance limitations due to inefficient state management and high communication overhead through external storage. We present GraphFlash, a fast and elastic graph processing framework built on serverless infrastructure. GraphFlash adopts a subgraph-centric programming model and leverages shared external storage for coordination and communication, enabling stateless, fine-grained function execution. It supports two execution modes: rotating mode for resource-constrained environments and pinned mode for higher performance when resources are sufficient. To address serverless limitations, GraphFlash introduces system-level optimizations, including partition-aware key aggregation, intra-function partition co-location, and superstep-aware activation. Across multiple graph algorithms and datasets, GraphFlash outperforms existing serverless-compatible systems by up to 127x in execution time and reduces resource consumption by up to 98% under higher-resource configurations, while matching the performance of traditional distributed frameworks on large workloads. Even with limited resources, it achieves up to 48x speedup and 99.97% cost reduction over prior serverless solutions, demonstrating that GraphFlash makes serverless graph processing practical and performant.

cs.DC↗
arXiv2026

Existent condition of partially wet state in capillary tubes

We develop a theory that predicts the equilibrium states of a fluid contained in a capillary which has corners. Each section of the tube can take three states: completely wet state where the tube section is completely occupied by the fluid, partially wet state where only the corners are occupied by the fluid known as corner film or finger, and completely dry state. We calculate the phase diagram of these states for a square tube with rounded corners. It is shown that the partially wet state can exist only in a certain region in the parameter space spanned by the equilibrium contact angle and the corner curvature.

cond-mat.soft↗
arXiv2026

Artificial-Noise-Aided Secure Near-Field MIMO With Fluid Antenna Systems

With the evolution of mobile communication systems toward large-scale arrays, high-frequency operation, and reconfigurable antenna architectures, fluid antenna systems (FAS) operating in the near-field (NF) regime provide new degrees of freedom (DoF) for secure and privacy-sensitive mobile access. This paper proposes an artificial-noise (AN)-aided physical layer security (PLS) scheme for NF fluid-antenna multiple-input multiple-output (FA-MIMO) systems, aiming to protect high-rate mobile service links supported by compact or large arrays. An alternating-optimization (AO) framework addresses the sparsity-constrained non-convex design by splitting it into a continuous BF/AN joint-design subproblem and a discrete FAS port-selection subproblem. Closed-form fully digital beamforming (BF)/AN solutions are obtained via a generalized spectral water-filling procedure within a block coordinate descent (BCD) surrogate and realized by a hardware-consistent hybrid beamforming (HBF) architecture with a shared RF network and independent digital BF/AN branches, while preserving the target BF/AN power split under constant-modulus RF constraints. For FAS port selection, a row-energy based prune--refit rule, aligned with Karush--Kuhn--Tucker (KKT) conditions of a group-sparsity surrogate, enables efficient active-port determination under a finite RF-chain budget. Simulation results confirm that the proposed design exploits the geometry and position-domain DoF of FAS and significantly improves secrecy performance, particularly for non-extremely-large arrays where NF beam focusing alone is inadequate. These results demonstrate the potential of AN-aided NF FA-MIMO as a practical secure-transmission architecture for future location-aware and hardware-constrained mobile computing systems.

cs.IT↗
arXiv2026

Correct Code, Vulnerable Dependencies: A Large Scale Measurement Study of LLM-Specified Library Versions

Large language models (LLMs) are now largely involved in software development workflows, and the code they generate routinely includes third-party library (TPL) imports annotated with specific version identifiers. These version choices can carry security and compatibility risks, yet they have not been systematically studied. We present the first large-scale measurement study of version-level risk in LLM-generated Python code, evaluating 10 LLMs on PinTrace, a curated benchmark of 1,000 Stack Overflow programming tasks. LLMs tend to specify version identifiers when directly prompted at 26.83%-95.18%, while down to 6.45%-59.19% in creating a manifest file directly. Among the specified versions, 36.70%-55.70% of tasks contain at least one known CVE, and 62.75%-74.51% of them carry Critical or High severity ratings. In 72.27%-91.37% of cases, the associated CVEs were publicly disclosed before the model's knowledge cutoff. The statistics show all models converge on the same small set of risky release versions, indicating a systemic bias rather than isolated model error. Static compatibility rates range from 19.70% to 63.20%, with installation failure as the dominant cause. The dynamic test cases confirm the pattern by 6.49%-48.62% pass rates. Further experiments confirm that these failures are attributable to version selection rather than code quality, and that externally anchored version constraints substantially reduce both vulnerability exposure and compatibility failures. Our findings reveal LLM version selection as a first-class, previously overlooked risk surface in LLM-based development. We disclosed these findings to the community of the evaluated models, and several confirmed the issue. All the code and dataset have been released for open science at https://github.com/dw763j/PinTrace.

cs.SE↗
arXiv2026

Rethinking Reasoning-Intensive Retrieval: Evaluating and Advancing Retrievers in Agentic Search Systems

Reasoning-intensive retrieval aims to surface evidence that supports downstream reasoning rather than merely matching topical similarity. This capability is increasingly important for agentic search systems, where retrievers must provide complementary evidence across iterative search and synthesis. However, existing work remains limited on both evaluation and training: benchmarks such as BRIGHT provide narrow gold sets and evaluate retrievers in isolation, while synthetic training corpora often optimize single-passage relevance rather than evidence portfolio construction. We introduce BRIGHT-Pro, an expert-annotated benchmark that expands each query with multi-aspect gold evidence and evaluates retrievers under both static and agentic search protocols. We further construct RTriever-Synth, an aspect-decomposed synthetic corpus that generates complementary positives and positive-conditioned hard negatives, and use it to LoRA fine-tune RTriever-4B from Qwen3-Embedding-4B. Experiments across lexical, general-purpose, and reasoning-intensive retrievers show that aspect-aware and agentic evaluation expose behaviors hidden by standard metrics, while RTriever-4B substantially improves over its base model.

cs.CL↗
arXiv2026

Guiding Self-Organizing Dynamics of Residential Choice in Cities to Reduce Traffic Congestion and Carbon Emissions

Rapid urbanization and growing vehicle ownership exacerbate traffic congestion and prolong commute times. We examine the self-organizing dynamics of residential choice via a hypothetical home-swapping process to mitigate peak-hour traffic congestion and carbon emissions. Specifically, we analyze over 400,000 trajectories from 9 days in a major Chinese city, revealing that actual average commuting distance is approximately three times shorter than under random residential distribution, indicating significant self-organization. Notably, city-wide home swapping reduces commuting distance by 50.4%, substantially easing traffic congestion, thereby reducing carbon emissions by 77.3%. Even with the consideration of socio-demographic factors and individual needs, the reductions remain significant: 8.1%-10.3% in commuting distance and 27.4%-34.4% in carbon emissions. Considering the potential induction of additional non-commuting trips, the reduction in carbon emissions remains substantial. Given the primacy of distance to the city center, polycentric city layouts can enhance these benefits. For validation, we use another dataset covering China's 28 major cities to confirm these findings. Finally, we introduce a data-driven model to elucidate self-organizing dynamics of residential choice and analyze the feasibility of government coordination. These insights demonstrate that a synergistic alignment of residential choices can leverage individual and city-level benefits, effectively alleviating commuting congestion and associated emissions.

physics.soc-ph↗
arXiv2026

Point Cloud Registration for Fusion between SPECT MPI and CTA Images

Clinical fusion of Single Photon Emission Computed Tomography Myocardial Perfusion Imaging (SPECT MPI) and Computed Tomography Angiography (CTA) remains limited by cross-modality misregistration and reliance on manual landmarks, which can hinder accurate ischemia localization and lesion-level functional assessment. To address this issue, we propose a registration and fusion framework for SPECT MPI and CTA that integrates functional and structural information for comprehensive cardiac evaluation. The proposed pipeline performs U-Net-based segmentation on both modalities. On SPECT MPI, only the left ventricle (LV) is extracted, and anatomical landmarks are automatically derived from characteristic LV structures. On CTA, both ventricles are segmented, and their spatial relationship is used to automatically define landmarks at the interventricular septal junction. Scale-space consistency preprocessing and landmark-driven coarse registration are applied to mitigate initial misalignment. Based on this initialization, multiple fine registration methods are evaluated on LV epicardial surface point clouds, including ICP, SICP, CPD, CluReg, FFD, and BCPD-plus-plus. The resulting transformations are then propagated to voxel-level resampling for high-precision SPECT-CTA fusion. In a retrospective cohort of 60 patients, the proposed framework preserved sub-millimeter coronary detail from CTA while accurately overlaying quantitative SPECT perfusion. Among the evaluated methods, BCPD-plus-plus achieved the highest accuracy with a mean point cloud distance of 1.7 mm. By combining robust initialization, comparative fine registration, and voxel-level fusion, the proposed approach provides a practical solution for myocardial ischemia localization and functional evaluation of coronary lesions, while remaining independent of any specific fine registration algorithm.

cs.CV↗
arXiv2026

Alice v1: Distillation-Enhanced Video Generation Surpassing Closed-Source Models

Wepresent Alice v1, a 14-billion parameter open-source video generation model that achieves state-of-the-art quality through consistency distillation with score regularization (rCM). Contrary to conventional distillation-which trades quality for speed-we demonstrate that rCM-based distillation can exceed teacher model quality. We attribute this to three mechanisms: (1) the score regularization term acts as a mode-seeking objective that concentrates probability mass on high-quality outputs rather than covering the full teacher distribution, (2) our targeted synthetic data pipeline with hard example mining provides training signal specifically for failure modes (physics, hands, faces) that the teacher handles inconsistently, and (3) consistency enforcement acts as implicit regularization, eliminating "lucky path" dependence on specific noise samples. Alice v1 generates 5-second 720p videos at 24fps in 4 denoising steps (~8 seconds on H100), a 7x speedup over the 50-step teacher while improving VBench score from 84.0 (Wan2.2) to 91.2. This surpasses both the teacher and closed-source systems including Veo3 (~90) and Sora2 (~88) on automated benchmarks, with competitive results in human preference studies. We release all model weights, training code, synthetic data pipelines, and evaluation scripts to advance open research in video generation.

cs.GR↗
arXiv2026

From Zero to Detail: A Progressive Spectral Decoupling Paradigm for UHD Image Restoration with New Benchmark

Ultra-high-definition (UHD) image restoration poses unique challenges due to the high spatial resolution, diverse content, and fine-grained structures present in UHD images. To address these issues, we introduce a progressive spectral decomposition for the restoration process, decomposing it into three stages: zero-frequency \textbf{enhancement}, low-frequency \textbf{restoration}, and high-frequency \textbf{refinement}. Based on this formulation, we propose a novel framework, \textbf{ERR}, which integrates three cooperative sub-networks: the zero-frequency enhancer (ZFE), the low-frequency restorer (LFR), and the high-frequency refiner (HFR). The ZFE incorporates global priors to learn holistic mappings, the LFR reconstructs the main content by focusing on coarse-scale information, and the HFR adopts our proposed frequency-windowed Kolmogorov-Arnold Network (FW-KAN) to recover fine textures and intricate details for high-fidelity restoration. To further advance research in UHD image restoration, we also construct a large-scale, high-quality benchmark dataset, \textbf{LSUHDIR}, comprising 82{,}126 UHD images with diverse scenes and rich content. Our proposed methods demonstrate superior performance across a range of UHD image restoration tasks, and extensive ablation studies confirm the contribution and necessity of each module. Project page: https://github.com/NJU-PCALab/ERR.

cs.CV↗