Search arXiv⌕ Search

arXiv subjects

Li Wang

Publications and source records attributed to Li Wang.

At least 19 recordsLinked to original sources
arXiv2026

Laws of the iterated logarithm for the range of once-reinforced random walk on $\mathbb{Z}$

In this paper, we study once-reinforced random walk on $\mathbb{Z}$, and prove sharp Chung-type and Hartman-Wintner laws of the iterated logarithm for the range. Let $R_n$ denote the range at time $n$ and let $c>0$ be the reinforcement parameter. Then, almost surely, \begin{equation*} \liminf_{n\to\infty}\frac{R_n}{\sqrt{n/\log\log n}}=\fracπ{\sqrt{2}}, \qquad \limsup_{n\to\infty}\frac{cR_n}{\sqrt{2n\log\log n}}=1. \end{equation*} The proofs are based on the independent increment decomposition of the inverse range and sharp generating function estimates.

math.PR↗
arXiv2026

Accurate structural modelling of protein-glycan recognition with GlycoWeave

Protein-glycan recognition underlies diverse biological processes, yet accurate prediction of protein-glycan complex structures remains challenging owing to the chemical diversity, conformational flexibility and sparse experimental structural coverage of glycans. Here we present GlycoWeave, a glycan-specific diffusion framework that integrates chemically constrained conformational sampling, multiscale molecular representations and multi-source learning. By operating in internal-coordinate space, GlycoWeave preserves glycan connectivity and stereochemistry while efficiently exploring bound conformations. A heterogeneous equivariant network couples atom-level interaction geometry with monosaccharide-level structural context, whereas multi-source, multi-head training integrates experimental complexes with structures refined using classical and quantum mechanics-informed potentials. On a benchmark of 698 protein-glycan complexes, GlycoWeave achieves a 71.9% Top-25 high-quality success rate, compared with 19.2% for AlphaFold 3, while requiring only 0.34 s per sampled pose. GlycoWeave further preserves glycan structural validity, recovers native protein-glycan interactions and reproduces linkage-specific conformational preferences, maintaining high accuracy across diverse glycan architectures and protein environments. These results establish GlycoWeave as an accurate, efficient and chemically faithful framework for modelling protein-glycan recognition. GlycoWeave is freely available at https://cadd.zju.edu.cn/glycoweave.

q-bio.BM↗
arXiv2026

Clustering Templates: A Data-Driven Framework for Dimensionality Reduction and Compression of Brain Strain in Head Impact

Purpose: Finite element (FE) models are indispensable for studying the biomechanics of traumatic brain injury (TBI), and modern head FE models typically comprise thousands to millions of elements. The resulting brain strain is therefore high-dimensional, and when combined with dense temporal sampling and large numbers of head impact cases, this produces datasets of substantial data volume. Accordingly, this study investigates the effective dimensionality required to accurately represent brain strain and proposes a data-driven framework to reduce the dimensionality of FE simulation data while maintaining high spatiotemporal fidelity. Methods: We developed a series of "clustering templates" by applying K-means clustering to feature vectors derived from the maximum principal strain (MPS) time-history curves of 118,840 brain elements over 3,693 impact simulations using the THUMS head model. The performance of these templates, with cluster counts (K) ranging from 8 to 8192, was rigorously evaluated by quantifying the fidelity of data reconstruction. Results: The clustering templates demonstrated high-fidelity reconstruction of the original FE data. A template with K = 512 achieved a compression ratio of over 200:1 while maintaining a low mean element-wise peak MPS root mean square error (RMSE) of 0.027, and a template with K = 4096 (30:1 compression ratio) achieved an RMSE of 0.016. Conclusion: The proposed clustering templates provide an effective and automated method for the dimensionality reduction and compression of large-scale brain strain data. This framework mitigates critical data management bottlenecks and enables the development of subsequent data analysis and machine learning models, thereby accelerating research in TBI biomechanics.

physics.bio-ph↗
arXiv2026

Stochastic Inverse Problem: stability, regularization and Wasserstein gradient flow

Inverse problems in physical or biological sciences often involve recovering an unknown parameter that is random. The sought-after quantity is a probability distribution of the unknown parameter, that produces data that aligns with measurements. Consequently, these problems are naturally framed as stochastic inverse problems. In this paper, we explore three aspects of this problem: direct inversion, variational formulation with regularization, and optimization via gradient flows, drawing parallels with deterministic inverse problems. A key difference from the deterministic case is the space in which we operate. Here, we work within a probability space rather than Euclidean or Sobolev spaces, making tools from measure transport theory necessary for the study. Our findings reveal that the choice of metric---both in the design of the loss function and in the optimization process---significantly impacts the stability and properties of the optimizer.

stat.ML↗
arXiv2026

RealComm: Benchmarking and Adapting Audio Deepfake Detection over Real Communication Channels

Acoustic capture, device processing, and transmission coding in real calls jointly alter speech signals and may affect the reliability of audio deepfake detection. To systematically evaluate these effects, we introduce RealComm, a dataset pairing digital speech with recordings of the same utterances transmitted through real mobile calls. The acquisition design covers acoustic and wired injection, handset and speakerphone operation, and caller-side denoising, with bona fide and synthetic speech recorded under matched conditions. The dataset includes RealCommBench for evaluation and RealCommTrain for adaptation. We further develop RealComm-Aug, a staged waveform augmentation pipeline for simulating acquisition and transmission effects, and investigate its combination with real-call training. Experiments with existing detectors reveal substantial degradation after call transmission, even for models with strong digital-domain performance; acoustic injection is particularly challenging. Paired quality and spectral analyses show that quality degradation is associated with poorer detection across recording conditions for the evaluated detectors, while high-frequency attenuation alone does not explain detection difficulty. Simulated augmentation and real-call training provide complementary benefits, with their combination achieving the lowest call-domain error for each of three evaluated detectors. For Teffic-Audio, the equal error rate decreases from 37.94% to 22.37%. The demo and benchmark access information are provided through the project repository.

eess.AS↗
arXiv2026

TempoMoE: Identifying and Steering Fast-Slow Thinking Experts for Efficient Multimodal Reasoning

Multimodal Mixture-of-Experts (MoE) reasoning models rely on long chain-of-thought (CoT) for complex tasks, yet often overthink easy inputs, increasing token cost and latency, and sometimes degrading accuracy. Existing methods optimize reasoning length during training or impose external control at inference, but leave the underlying reasoning mechanism a black box. We find that fast and slow thinking exhibit distinct expert-routing patterns: holding the input fixed and varying only the prompted reasoning tempo produces stable routing shifts. Suppressing slow-thinking experts shortens the reasoning chain, whereas suppressing fast-thinking experts lengthens it, providing causal evidence that these experts influence reasoning depth. Based on this finding, we propose TempoMoE, a training-free, two-stage framework. It identifies fast- and slow-thinking experts via tempo-contrastive routing analysis and redistributes routing weights within the original top-k set according to sample difficulty. It amplifies fast-thinking experts and suppresses slow-thinking experts on easy inputs and reverses this adjustment on hard ones, enabling difficulty-adaptive control over reasoning depth. We evaluate TempoMoE on Kimi-VL, Qwen3-VL, and ERNIE-4.5-VL across six benchmarks. Compared with Vanilla, TempoMoE consistently improves accuracy while reducing output tokens, with accuracy gains of up to 2.6 points and token reductions of up to 35.4%.

cs.CL↗
arXiv2026

Right Values, Wrong Places: Diagnosing and Fixing Record Misplacement in LLM-Generated JSON

Large language models can generate valid JSON that preserves field values yet binds them to the wrong records or hierarchy levels. We introduce JSONPlacement, a controlled task supplying all values and assignment rules for transforming flat records into hierarchical JSON. Across six open-weight models, we identify two error types: record-assignment errors and hierarchy-realization errors. Whole-record moves account for 98.6% of strictly traceable synthetic misplacements. We predict that aligning input with target traversal should reduce assignment errors, while replacing nested output with record-level paths should reduce hierarchy errors. With predictions fixed before generating new instances, we test both interventions separately and together. On synthetic instances, each nearly eliminates its targeted error while leaving the other. Combined, they achieve exact matches for three 8-12B models on all new synthetic instances and a real-data validation set. These findings motivate evaluating record placement explicitly and jointly designing input organization and output representation to preserve value-path bindings.

cs.AI↗
arXiv2026

Learning What to Keep: Transfer-Aware Personal Skill Learning from User Feedback

User feedback enables LLMs to refine reusable personalized skills, but often mixes persistent preferences with request-specific corrections. Indiscriminately retaining both risks negative transfer, motivating selective retention at the level of individual requirements. We frame this challenge as selective transfer over atomic rules and propose SkillGrain, a framework for learning reusable personalized skills from user feedback. SkillGrain decomposes feedback into independently actionable rules, predicts their transferability, and retains only those suitable for future requests. To learn these decisions, we combine rubric-guided skill rewards with fine-grained credit assignment that separately evaluates rule content and retention decisions. We further jointly optimize skill induction and skill-conditioned generation through reinforcement learning, ensuring that learned skills support downstream personalization. We construct a benchmark spanning five personalized writing tasks to evaluate cross-request generalization. Experiments across three base models in both in-domain and out-of-domain settings show that SkillGrain achieves the best overall average performance among competing methods.

cs.CL↗
arXiv2026

CoReLoop: Parameter-Efficient Controlled Recurrent Refinement for Audio Deepfake Detection

Audio deepfake detectors are expected to remain robust to diverse attacks, datasets, and recording conditions, yet collecting training data that cover every future condition is impractical. We investigate whether an already-trained SSL-based detector can improve its generalization through additional recurrent computation, without introducing new training data or updating its original parameters. Directly feeding the encoder output back as its next input, however, severely degrades detection, revealing a mismatch between the representations produced and consumed by a model trained for single-pass inference. We therefore propose CoReLoop, which enables controlled recurrent refinement through loop-specific LoRA and lightweight modules that adapt recurrent inputs, regulate state updates, and align refined representations with the frozen classifier. Across 14 diverse test sets, CoReLoop improves pooled EER across 3-, 6-, 12-, and 24-layer encoders; on the 24-layer model, two-pass refinement reduces EER from 4.85% to 3.74% while training only approximately 11M of 598M parameters. An optional halting head further selects refinement depth per utterance, achieving 3.73% EER with an average of 1.18 passes. Analysis shows that later-pass saturation coincides with smaller applied updates and nearly balanced corrections and regressions, while the recurrent state continues to change.

cs.SD↗
arXiv2026

CIRRA: Dual-Level Continual Instruction Reconciliation with Ongoing Execution for Embodied Robot Agents in Interactive Household Tasks

Household robots must accommodate new user instructions while executing ongoing tasks. Existing agents often regenerate or extensively revise the remaining task sequence, introducing plan ambiguity, logical inconsistency, and redundant execution. We formulate continual instruction reconciliation and propose CIRRA (Continual Instruction Reconciliation for Robot Agents), a dual-level framework combining LLM-based semantic reasoning with rule-constrained structural integration. CIRRA first grounds incoming instructions to unique executable skills and resolves underspecified actions and execution locations. It then preserves the ongoing subtask sequence as an execution backbone and generates integration candidates by inserting incoming subtasks into location-matched segments. The semantic reasoner evaluates only modified segments to identify dependencies and conflicts and select the most logically coherent local integration. This structure-preserving process maintains alignment with ongoing execution, mitigates ambiguity and inconsistency, and reuses shared subtasks to reduce redundant execution. We also introduce CHIRP (Continual Household Instruction Reconciliation and Planning), a text-based benchmark of 120 episodes across eight household environments and six categories of everyday activities. On CHIRP, CIRRA achieves 74.2% decision agreement, exceeding the strongest replanning baseline by 30 percentage points; every correct fusion decision yields a correctly placed, conflict-free schedule. On a Unitree G1 humanoid, CIRRA interrupts ongoing skills at the correct moment in every trial and significantly outperforms all baselines on every metric.

cs.RO↗
arXiv2026

Fundamental solution expansions in stochastic homogenization of parabolic systems with time-dependent coefficients

This paper studies fundamental solutions of uniformly parabolic systems with stationary random coefficients depending on space and time. Under ergodicity and local regularity, we establish quenched Gaussian bounds for the kernel and its spatial derivatives, with stationary random prefactors. A spectral gap assumption gives moments of every finite order for these prefactors and hence annealed Gaussian bounds. We also prove quantitative two-scale expansions for the kernel, its spatial gradient, and its mixed spatial derivative. The error estimates hold in every finite probability moment and retain Gaussian decay. Their dimension-dependent growth is described by a corrector weight, with a single additional logarithmic factor in the derivative estimates. The proofs combine large-scale regularity, extended correctors, weighted estimates, and a parabolic weak-norm estimate. The latter supplies the passage from deterministic pairings to local annealed estimates without interchanging expectation and a supremum over test functions. The mixed-derivative estimate builds on the gradient expansion and does not introduce a further logarithmic loss.

math.AP↗
arXiv2026

DS-Net: A Dynamic Split Network for Edge-Cloud Collaborative Inference

Collaborative inference serves as a critical methodology in edge intelligence. Existing collaborative inference methods operate by splitting the deep neural network (DNN) model between edge devices and cloud servers. However, these methods need to load different models for multiple split points when the communication or device changes, thus resulting in long switching times and substantial storage overhead on the edge side. To address this issue, we propose a Dynamic Split Network (DS-Net), which not only enables efficient split point switching but also enhances the collaborative speed and accuracy across various scenarios. Specifically, we design a dynamic switching framework that enables flexible selection among multiple split points within a single model, thereby reducing storage, memory, and switching overhead. To mitigate the feature distribution shifts and task-relevant information loss caused by dynamic splitting and high-ratio compression, we further propose a hierarchical compensation mechanism. Local alignment calibrates the decoded feature distribution at each split point, while task-specific global compensation restores task-relevant information at the global level. In addition, we incorporate sample-adaptive global compensation selection into split-point selection based on sample complexity and runtime resource conditions. Experimental results demonstrate that DS-Net reduces memory overhead by 74.6% and achieves millisecond-level switching latency. Moreover, under a 10 Mbps bandwidth constraint on the Nano platform, DS-Net outperforms state-of-the-art collaborative methods by achieving 1.2% higher accuracy on the COCO dataset, reducing latency by 42.5%, and improving throughput by a factor of 1.8$\times$.

cs.LG↗
arXiv2026

Ideal balanced pairs in n-exangulated categories

This paper is devoted to ideal approximation theory in $n$-exangulated categories. We define ideal balanced pairs and explore their properties and equivalent characterizations. We prove that any complete ideal cotorsion triple $(\mathcal{I},\mathcal{K},\mathcal{J})$ satisfying suitable assumptions yields an ideal balanced pair $(\mathcal{I}, \mathcal{J})$, and the converse holds under additional hypotheses. We further show that for an ideal balanced pair, it is equivalent to an additive subfunctor with enough projective morphisms and enough injective morphisms.

math.CT↗
arXiv2026

CheckerBench: Can Long-Horizon Agents Synthesize Static-Analysis Checkers?

Static-analysis checker synthesis requires agents to interpret a defect specification, inspect a repository, implement analyzer-specific logic, and refine the checker through repeated compilation and analysis feedback. Existing coding-agent benchmarks focus on tasks such as patch generation or vulnerability detection and rarely assess whether an agent can develop a working checker in a repository from start to finish. We introduce CheckerBench, an executable benchmark of 300 tasks derived from 297 CVEs across 167 repositories, 85 CWEs, and five language ecosystems. Each task includes vulnerable and fixed revisions, a pinned analysis environment, and a checker scaffold. We further introduce CheckerLab, a common evaluation framework that independently rebuilds submitted checkers and measures vulnerable-fixed diagnostic contrast, patch localization, false positives, and tool use. Across 21 model-harness configurations and three independent repeats per configuration, mean Pass@1 is 32.30%, while the best reaches 45.33%. These results show that reliable, reusable checker development remains challenging for current coding agents.

cs.SE↗
arXiv2026

Sparse2comm: Towards Robust Cooperative 3D Object Detection

Cooperative perception improves autonomous driving by sharing complementary observations among vehicles and roadside infrastructure for 3D object detection. However, practical deployment is constrained by limited bandwidth and unreliable cooperation, where packet loss, transmission delay, and spatial misalignment jointly degrade the cooperative feature stream. Existing methods often reduce communication cost or compensate for one degradation type, leaving coupled disturbances insufficiently addressed. To address this problem, we propose Sparse2comm, a bandwidth-efficient and robust cooperative 3D object detection framework that treats unreliable cooperation as progressive restoration over degraded cooperative features. Sparse Feature Encoding first encodes communication as randomly mask-sampled foreground features transmitted by collaborating agents, from which the ego vehicle reconstructs dense semantic representations. This sparse-to-dense mechanism learns to infer missing object-centric content from sparse observations, enabling ultra-low-bandwidth communication and packet-loss recovery within the same representation. On the semantically restored features, Latency-Aware Alignment predicts motion flow to compensate delayed messages, and Self-Calibrating Fusion estimates residual spatial offsets in a self-supervised manner before adaptive cross-agent fusion. Sparse2comm therefore restores semantic completeness, temporal consistency, and spatial alignment in an ordered pipeline. Extensive experiments on DAIR-V2X, OpenV2V, and V2V4Real show that Sparse2comm maintains competitive clean accuracy and consistently improves robustness under individual and mixed real-world degradations. Compared with the selective feature communication baseline Where2comm, Sparse2comm improves mixed-setting AP@0.5/AP@0.7 by +20.15/+11.79, +12.66/+11.07, and +15.36/+12.61 on the three datasets, respectively.

cs.CV↗
arXiv2026

Defenses at Odds: Measuring and Explaining Defense Conflicts in Large Language Models

Large language models (LLMs) may require additional defenses after deployment as risks and governance requirements evolve. Subsequent defenses can interact with earlier defenses, raising the question of their sequential compatibility. We study this question with CONFLICTEVAL, evaluating 144 ordered compositions of six defenses spanning safety, privacy, and fairness across six models. The resulting interactions are heterogeneous across defense pairs, application orders, and models. Notably, some compositions exhibit defense conflicts: the subsequent defense improves its target objective while weakening protection established by the earlier defense. We investigate these interactions through the directional compatibility of defense-induced changes in risk-relevant representations. Across 11 selected cases, we use activation interventions to assess which defense-induced representational changes support protection and examine how subsequent defenses affect these changes. We find that, in some conflict cases, subsequent defenses counteract representational changes supporting earlier protection, providing evidence for one possible pathway to defense conflicts. Building on this analysis, we propose Conflict-Triggered Directional Retention (CTDR), which penalizes opposing shifts along directions supporting earlier protection. On six selected conflicting compositions from this analysis, CTDR reduces first-objective regression while maintaining positive gains on the subsequent defense objective.

cs.CR↗
arXiv2026

MomADv2: Reliable Temporal Memory for End-to-End Autonomous Driving

Long-horizon planning is critical for safe autonomous driving in complex scenarios. Existing methods improve planning continuity with temporal memory, but such memory may become invalid and mislead decisions when the driving command changes. Thus, selectively leveraging useful history while suppressing command-inconsistent memory remains a key challenge. To address this issue, we propose MomADv2, a reliable state-space memory framework for long-horizon end-to-end autonomous driving. At its core, MomADv2 introduces a Selective State-Space Planning Memory Query Module, which filters historical planning queries based on temporal continuity and command consistency, selects planning modes relevant to the current command, and models the evolution of planning intentions through a selective state-space mechanism. To further alleviate local trajectory deviations and error accumulation in long-horizon planning, we design a Flow-Matching Trajectory Residual Refiner. It learns a continuous residual correction field from the refined planning output to the expert trajectory, enabling fine-grained trajectory refinement while preserving the stability of anchor-based planning. Extensive experiments on closed-loop NAVSIM and Bench2Drive, as well as open-loop nuScenes, demonstrate that MomADv2 improves long-horizon planning consistency and reduces the average collision rate by 15.6% over MomAD under 6-second planning.

cs.CV↗
arXiv2026

SceneJail: Exploiting Video Scenario Context to Jailbreak Multimodal LLMs

Video Multimodal Large Language Models (Video-MLLMs) support reasoning over video inputs, yet remain vulnerable to jailbreak attacks that elicit policy-violating responses. Existing video jailbreaks primarily manipulate how harmful queries are visually presented, thereby treating video merely as a carrier. Consequently, the surrounding video scenario remains unexplored as a contextual attack surface. In this paper, we show that the same harmful query can elicit different safety responses when placed in different video scenarios. To systematically exploit this vulnerability, we propose SceneJail, an adaptive black-box jailbreak framework with two coordinated components. Adaptive Scenario Construction dynamically searches for a surrounding scenario that is contextually compatible with the harmful query. Scenario-aware Prompt Search uses black-box response feedback to search for textual guidance tailored to the selected scenario. Extensive evaluations on the HADES and SafeBench datasets across eight Video-MLLMs, including two proprietary models, GPT-4.1 and Gemini3.5-Flash, demonstrate the effectiveness of SceneJail. SceneJail-F, which presents the complete query persistently, achieves average attack success rates (ASR) up to 91.5%, outperforming the strongest baselines by 29.1 percentage points. Furthermore, SceneJail-S, which distributes the query across successive frames, remains highly robust against current defenses, retaining a 72.3% ASR even under strict image filtering.

cs.CR↗