arXiv subjects

Wei Chen

Publications and source records attributed to Wei Chen.

At least 109 records · Page 6Linked to original sources

arXiv2026

Localization phase diagram of the Hexagonal Lattice with irrational magnetic flux

We study the Hofstadter model on a hexagonal lattice with irrational magnetic flux in this work. The Hofstadter model of the square lattice with irrational flux has been solved mathematically by Avila and his collaborators in his Fields medal work. However, this theory is usually not applicable to lattices with internal degrees of freedom, such as spin or sublattice. In this work, we show that for the hexagonal lattice with only nearest neighbor hopping, the system can still be characterized by a two by two transfer matrix and solved exactly by the Avila global theory although this lattice has two sublattices. We obtained the exact localization phase diagram of the hexagonal lattice with irrational flux by this theory, which reveals three pure phases, i.e., the extended, localized and critical states but no mobility edge due to the chiral symmetry. We used the renormalization group (RG) theory to verify these results, which can determine part of the phase diagram. We then computed the fractal dimension of the remaining part numerically. The results from both the RG theory and numerical analysis confirmed the phase diagram we get from the Avila global theory. Our results can be tested in various hexagonal Moire lattices and artificial superlattices in recent experiments.

cond-mat.mes-hall↗

arXiv2026

Enhancing Large Language Model Reasoning with Reward Models: An Analytical Survey

Reward models (RMs) play a critical role in enhancing the reasoning performance of LLMs. For example, they can provide training signals to finetune LLMs during reinforcement learning (RL) and help select the best answer from multiple candidates during inference. In this paper, we provide a systematic introduction to RMs, along with a comprehensive survey of their applications in LLM reasoning. We first review fundamental concepts of RMs, including their architectures, training methodologies, and evaluation techniques. Then, we explore their key applications: (1) guiding generation and selecting optimal outputs during LLM inference, (2) facilitating data synthesis and iterative self-improvement for LLMs, and (3) providing training signals in RL-based finetuning. Finally, we discuss critical open questions regarding the selection, generalization, evaluation, and enhancement of RMs, based on existing research and our own empirical findings. Our analysis aims to provide actionable insights for the effective deployment and advancement of RMs for LLM reasoning.

cs.CL↗

arXiv2026

Safety-Aware Forward Detection in Networked ISAC for Low-Altitude UAV Flight

Networked integrated sensing and communication (ISAC) exploits cooperation among multiple ground base stations (GBSs) to support safe uncrewed aerial vehicle (UAV) flight in low-altitude wireless networks (LAWNs). Existing studies mainly focus on communication enhancement or target parameter estimation, while the detection reliability of non-cooperative targets in the UAV forward region remains insufficiently investigated. To address this issue, this paper proposes a safety-aware forward detection design in networked ISAC, where multiple GBSs jointly support UAV downlink communication, state estimation, and non-cooperative target detection within the forward region of interest (ROI). First, the forward ROI is determined by the UAV position, velocity, and safe braking distance, and is voxelized to characterize target-existence states. Then, the Cramér-Rao lower bound (CRLB) for UAV state estimation and the forward-ROI miss-detection probability are derived, and their scaling laws are characterized: In detail, the UAV state-estimation CRLB approximately decreases as $\ln^{-2}J$ with the number of cooperative GBSs $J$, while the forward-ROI miss-detection probability follows an exponential-form scaling law as $λ_{t}D_{f}\ln^{-2}J$. Furthermore, a safety-aware resource optimization problem is formulated to jointly configure the sensing pilot ratio, transmit power, and beam direction, balancing UAV state-estimation performance and forward detection reliability under the communication-rate constraint. Simulation results show that, compared with the baseline scheme without forward detection, the proposed design reduces the average miss-detection probability and the corresponding sensing-induced collision risk by $17.05\%$, while introducing only limited state-estimation performance degradation, reflected by a $14.82\%$ increase in the average CRLB.

cs.IT↗

arXiv2026

TALSC: Timeliness-Aware Large-Small VLM Collaboration for Infrastructure-Assisted Autonomous Driving

The deployment of Vision-Language Models (VLMs) in autonomous driving (AD) systems is constrained by on-board computing power, restricting vehicles to small VLMs (SVLMs) with limited perception and reasoning capabilities. Infrastructure-assisted AD alleviates this resource constraint by enabling collaboration with large VLMs (LVLMs) at edge servers. However, in dynamic vehicular environments, the utility of sensory data for downstream tasks decays rapidly, making timeliness of information a critical concern. To balance the accuracy gains of LVLMs with their latency-induced timeliness degradation, we develop a Timeliness-Aware Large-Small VLM Collaboration (TALSC) framework. Specifically, we first model the Age of Information (AoI) evolution for VLM inference and characterize the coupling among AoI, token length, and task performance to formulate a general timeliness metric. Building on this, we propose the TALSC online scheduling algorithm. Since scheduling decisions have a delayed impact on future timeliness metric and the output token number is unknown at scheduling time, we design a Lyapunov drift-plus-estimated-penalty algorithm and provides a guaranteed performance. In simulation, we first conduct a case study to derive a fitted timeliness metric based on nuScenes dataset, and further show that TALSC outperforms baselines under various communication and computing settings, achieving up to a 12.6\% normalized improvement in Micro-F1 score compared with the best-performing baseline.

cs.DC↗

arXiv2026

Unpaired Modality-Agnostic Generative Recommendation

Generative Recommendation (GR) formulates recommendation as autoregressive generation over discrete semantic identifiers (IDs). Although recent multimodal GR methods improve semantic ID construction with visual and textual information, they typically require item-level paired observations, restricting tokenization to the intersection of modality availability. Moreover, incorporating unpaired observations is nontrivial because small representation shifts may cross quantization boundaries and produce incompatible identifier sequences. To address this challenge, we propose \textbf{Unpair}ed Modality-Agnostic \textbf{G}enerative \textbf{R}ecommendation (UnpairGR), which learns a unified semantic-ID space from paired, image-only, and text-only observations. UnpairGR confines modality-specific processing to lightweight input projections while sharing the subsequent Transformer and residual codebooks across all observation conditions. Paired observations establish a reliability-guided cross-modal consensus, whereas unimodal observations directly refine the same representations and codes. The learned tokenizer is then fixed to provide stationary targets for a single autoregressive recommender, without feature imputation, modality-specific codebooks, or fallback mappings. Extensive experiments on three benchmark datasets demonstrate that UnpairGR consistently improves recommendation performance under both fully observed and incomplete-observation settings.

cs.IR↗

arXiv2026

Requirement--Evidence Alignment for Compositional E-Commerce Queries

Compositional e-commerce queries express multiple requirements that must hold jointly, yet existing rerankers collapse these constraints into aggregate relevance and often promote topical near misses over feasible products. In this paper, we introduce REAlign, a novel requirement-evidence-aligned reranking framework that explicitly connects typed query requirements with visible evidence. REAlign distinguishes satisfied, violated, and unsupported conditions, constructs requirement-targeted contrasts that expose failure modes, and optimizes duplicate-free partial rankings through Requirement-Aware Group-Relative Policy Optimization. Its list utility preserves relevance while incorporating requirement satisfaction, evidence support, material violations, and output validity. Experiments on two fixed-pool e-commerce benchmarks show consistent improvements over strong supervised and policy-optimization baselines under matched training budgets, with fewer violations among top-ranked candidates and larger gains at shallow ranks. Controlled ablations confirm the complementary value of requirement modeling, evidence grounding, and decomposed optimization.

cs.IR↗

arXiv2026

Local-Time Riemannian Score Matching on the Quantum Pure-State Manifold

Score-based diffusion can be defined intrinsically on the manifold of quantum pure states, $\mathbb{CP}^{d-1}$ with the Fubini--Study metric, but no closed-form transition density is available, so the score must be supervised by a local-time teacher taken from the Euclidean limit of the diffusion in normal coordinates. This paper is about what makes that teacher work, and where it stops working. Three training choices turn out not to be incidental: the increment must be divided by the diffusion clock rather than by the elapsed time, since the published expression assumes unit diffusion and a non-unit schedule introduces a time-change mismatch varying by a factor of $400$ across the horizon; the logarithm and exponential maps should be the closed-form Fubini--Study ones, which is the largest single effect we measure; and the global phase must be randomised, because horizontal projection alone does not make a score network descend to the quotient. With these choices the model improves on the published Riemannian local-time baseline in every cell of an eight-benchmark, four-metric comparison over ten seeds, significantly on five of eight after Holm correction, and beats an ambient Euclidean baseline by an order of magnitude everywhere. We then bound what the approximation costs by replacing it with the exact heat kernel of $\mathbb{CP}^{d-1}$, computable up to complex dimension seven, where the local-time teacher loses a factor of $1.2$ to $2.7$.

stat.ML↗

arXiv2026

Control variates with neural surrogates for uncertainty quantification in kinetic equations

Efficient uncertainty quantification for kinetic equations with random inputs is challenging because it requires repeated simulations of high-dimensional models, such as the Boltzmann, Landau, and related collisional equations, whose computational cost can quickly become prohibitive. Multifidelity control variates address this difficulty by coupling a small number of high-fidelity simulations with many evaluations of lower-complexity reduced models. In this work, we analyze the case in which the reduced model is replaced by a neural surrogate rather than evaluated through a classical numerical scheme. We show that the resulting estimator remains unbiased and that the change in the optimal variance induced by the neural approximation is controlled by the error between the exact low-fidelity observable and its neural approximation. This estimate is then combined with residual stability estimates for inhomogeneous Fokker--Planck and Bhatnagar--Gross--Krook surrogates. We also extend the analysis to several control variates and to an asymptotic-preserving (AP) hierarchy containing the limiting Euler observable. In the fluid limit, the optimal hierarchical variance converges to the variance associated with the limiting Euler control, while the contribution of the intermediate kinetic correction vanishes. Numerical tests based on micro--macro neural surrogates illustrate the predicted variance stability and the behavior of the AP hierarchy.

math.NA↗

arXiv2026

Breaking the Exploration Bottleneck: Rubric-Scaffolded Reinforcement Learning for General LLM Reasoning

Recent advances in Large Language Models (LLMs) have underscored the potential of Reinforcement Learning (RL) to facilitate the emergence of reasoning capabilities. Despite the encouraging results, a fundamental dilemma persists as RL improvement relies on learning from high-quality samples, yet the exploration for such samples remains bounded by the inherent limitations of LLMs. This, in effect, creates an undesirable cycle in which what cannot be explored cannot be learned. In this work, we propose Rubric-Scaffolded Reinforcement Learning (RuscaRL), a novel instructional scaffolding framework designed to break the exploration bottleneck for open-ended reasoning. Specifically, RuscaRL introduces checklist-style rubrics as (1) explicit scaffolding for exploration during rollout generation, where different rubrics are provided as external guidance within task instructions to steer diverse high-quality responses. This guidance is gradually decayed over time, encouraging the model to internalize the underlying reasoning patterns; (2) verifiable rewards for exploitation during model training, where we can obtain robust LLM-as-a-Judge scores using rubrics as references, enabling effective RL on open-ended reasoning tasks. Extensive experiments demonstrate the superiority of the proposed RuscaRL across various benchmarks, effectively expanding reasoning boundaries under the Best-of-N evaluation. Our code is available at https://github.com/IANNXANG/RuscaRL.

cs.LG↗

arXiv2026

Harnessing the Wisdom of LLM Crowds through Complementarity-Driven Iterative Collaboration

Large language models (LLMs) are increasingly deployed in enterprise settings, yet individual models remain bounded by model-specific capability limitations. These heterogeneous boundaries pose a deployment challenge, but also create an opportunity: strategically coordinating multiple LLMs may unlock collective intelligence exceeding any single model. Existing approaches fix how models are combined in advance, overlooking the dynamic, state-dependent role of complementarity in complex problem solving. Drawing on the wisdom-of-crowds paradigm, we reconceptualize collective LLM intelligence as relay-style complementarity: a sequential process in which each successor model is selected to address the specific bottleneck identified in its predecessor's output. To operationalize this, we propose WILC (Wisdom Integration of LLM Crowds), a framework grounded in two design principles. First, iterative reflection-and-refinement establishes a state-preserving workflow through which models diagnose and refine prior outputs. Second, complementarity-driven model selection governs transitions via a dual-gate mechanism: prospective complementarity fit (PCF) identifies the worker most suited to the current bottleneck, while posterior complementarity gain (PCG) evaluates whether the selected transition improves the evolving solution. Experiments across four diverse benchmarks show that WILC outperforms existing approaches, including single-model self-refinement, ensemble methods, and query-routing methods. Under standardized pricing assumptions, WILC matches the average benchmark performance of GPT-5.2 at roughly 7 times lower estimated per-query cost, while facilitating data sovereignty through self-hosted deployment. This study extends wisdom-of-crowds theory from static aggregation to sequential AI complementarity and provides transferable design principles for multi-AI coordination.

cs.AI↗

arXiv2026

Prox: Training-Free FFN Activation Sparsity via Approximate Intermediate-Channel Salience in LLMs

Feed-forward networks (FFNs) dominate memory traffic and computation in large language model (LLM) inference, making them a primary target for activation sparsification. However, existing training-free methods suffer substantial model-quality degradation at high sparsity due to limitations in their channel-selection strategies. We observe that the SwiGLU intermediate state provides a highly effective channel-selection signal, but obtaining it requires costly dense computation. To address this, we present \emph{Prox}, a two-stage training-free framework for sparse SwiGLU FFNs. Prox hinges on the key insight: sparse execution requires only the channel mask induced by the intermediate state, which can be constructed from the magnitude ranking of its entries rather than their exact values. Specifically, Stage 1 uses input sparsity and quantized proxy weights to construct a shared mask; Stage 2 computes the selected channels exactly, enabling sparse execution of all three projections. Across ten LLMs from six model families, Prox outperforms training-free baselines at all sparsity levels, achieves up to a $1.99\times$ end-to-end decoding speedup at 70\% FFN sparsity, and is compatible with quantization and sparse attention.

cs.LG↗

arXiv2026

Semantic-Aligned Structural Abstraction for Multimodal Sentiment Analysis

Multimodal Sentiment Analysis (MSA) aims to interpret complex human emotions by integrating natural language with non-verbal modalities. Non-verbal modalities share a structural isomorphism with natural language, as both can be viewed as feature sequences evolving over time. This isomorphism enables the transformation of non-verbal modalities into text-like tokens for unified semantic reasoning. Large Language Models (LLMs), designed to understand and generate sequential data, can thus be utilized to interpret complex affective sequences. However, existing LLM-based methods primarily capture low-level superficial features, failing to model affective semantics arising from structural variations and contextual interactions. To address this limitation, we propose \textbf{SentiLLM}, a unified framework that leverages \textit{Semantic-Aligned Structural Abstraction} to distill continuous raw signals into compact, semantically meaningful tokens. Specifically, we introduce a \textit{Dual-Stream Salience-Context Calibration Mechanism}, which disentangles non-verbal feature sequences into a focus stream and an ambient stream. The focus stream captures salient sentiment shifts (e.g., facial expressions) guided by textual priors, while the ambient stream characterizes stable background states. Through calibrating these dynamic sentiment shifts against background states, SentiLLM effectively projects non-verbal modalities into a unified semantic space, making them naturally understandable for LLMs. Serving as a plug-and-play module, SentiLLM significantly improves discriminative performance with only a small number of trainable parameters. Our method achieves superior performance on four datasets, MOSI, MOSEI, CH-SIMS, and CH-SIMS v2, demonstrating the effectiveness of the structural abstraction paradigm in MSA. Our code is available at: \href{https://github.com/especiallyW/SentiLLM}.

cs.CL↗

arXiv2026

AniMaster: From Story Texts to Animated Videos via Cinematic Script Generation and Interactive Authoring

Recent advances in Video Generation Models (VGMs) have demonstrated strong capabilities in producing short video clips. However, it is still challenging for everyday creators to leverage these models to produce polished long-form animated videos from brief story texts. Informed by a formative study with both novice creators and film experts, we identify two major challenges of interactive video authoring: (1) the lack of expertise in translating free-form story texts to professional cinematic scripts and finally high-quality animated videos, and (2) the absence of effective ways to convey video design intents to key variables of visual storytelling, such as shot composition, camera controls and shot sequencing. Drawing on narratology and film studies, we propose a three-layer design framework that defines the key design dimensions across three layers (i.e., story texts, cinematic scripts, and animated videos) as well as the translation between them. Built on this framework, we present AniMaster, a VGM-powered authoring tool to enable everyday creators to easily produce smooth animated videos from free-form story texts. AniMaster automatically expands brief story texts to detailed cinematic scripts, and further translates cinematic scripts into polished videos by following professional visual storytelling principles. It also allows users to interactively edit the scripts and refine the generated videos via text instructions and intuitive interactions. We extensively evaluated AniMaster through an in-depth user study with 16 participants, two case studies, and expert interviews with 2 film professionals. The results demonstrate the effectiveness and usability of AniMaster in helping everyday creators create polished animated videos from free-form story texts.

cs.HC↗

arXiv2026

Demystifying Deep Learning Compiler Frontend Bugs: An LLM-Aided Empirical Study

Deep learning compilers (DLCs) are designed to translate deep learning programs into optimized, hardware-specific code. Typically, DLC frontends translate programs into graph-based intermediate representations (IRs) to enable optimizations. Defects introduced during this stage (termed \emph{fBug}s) are severe yet understudied, as prior work predominantly focuses on low-level APIs and operators or treats DLCs as monolithic entities. To bridge this gap, we conduct the first systematic empirical study of \emph{fBug}s in TorchDynamo, the default DLC frontend for PyTorch 2, the most popular DL framework. Leveraging a domain-knowledge-enhanced LLM-aided methodology, we analyze 123 \emph{fBug}s and construct a taxonomy comprising 7 root cause categories and 15 subcategories. Our findings provide actionable insights for DLC development and testing. Furthermore, we leverage the LLM to generate targeted, root cause-aware test cases to detect new bugs. We uncovered 23 previously unknown \emph{fBug}s in recent releases (15 confirmed) across eight (sub)categories, demonstrating the efficacy of our methodology in testing and hardening DLC frontends.

cs.PL↗

arXiv2026

LiteResearcher: A Scalable Agentic RL Training Framework for Deep Research Agent

Reinforcement Learning (RL) has emerged as a powerful training paradigm for LLM-based agents. However, scaling agentic RL for deep research remains constrained by two coupled challenges: hand-crafted synthetic data fails to elicit genuine real-world search capabilities, and real-world search dependency during RL training introduces instability and prohibitive cost, which limits the scalability of Agentic RL. LiteResearcher is a training framework that makes Agentic RL scalable: by constructing a lite virtual world that mirrors real-world search dynamics, we enable a continuously improving training recipe that empowers a tiny search agent to outperform large-scale open-source and commercial models (e.g., Tongyi DeepResearch and Claude-4.5 Sonnet). Specifically, on common benchmarks such as GAIA and Xbench, our LiteResearcher-4B achieves open-source state-of-the-art results of 71.3% and 78.0% respectively, demonstrating that scalable RL training is a key enabler for Deep Research Agents.

cs.AI↗

arXiv2026

Side-channel-secure quantum key distribution with correlated sources

Quantum key distribution (QKD) offers theoretical security guarantees for sharing secure key, but its practical systems face challenges due to the imperfections of devices. Widespread quantum state preparation imperfections, such as correlations between multiple rounds, significantly undermine the real-world security of QKD. In this paper, we propose a protocol that is immune to almost all kinds of state-preparation imperfections over multiple correlated rounds arising from both encoding and unknown non-encoding dimensions. The protocol relies only on three assumptions: the imperfect encoding produces unknown product states rather than entangled ones, a lower bound on the vacuum components is known, and the correlation has a finite range. The proposed protocol is also measurement-device-independent, ensuring high security at both the source and measurement sides. We provide the finite-key security analysis against coherent attacks and conduct numerical simulations to see the performance. The results show that for small correlation ranges, the protocol achieves excellent performance with a maximal transmission loss exceeding 60 dB (>300 km in standard fiber). Even for extreme cases, where one encoding affects up to 500 neighboring rounds, the protocol can still generate secret keys over a 10 dB-loss channel.

quant-ph↗

arXiv2026

Thermo-Structural Topology Optimization Considering Nonlinear Creep

Creep is a primary life-limiting mechanism for metallic components operating at high temperature, producing permanent deformation under sustained loads even when stresses remain below yield. The design of structures to minimize this deformation is critical to extending the service life of components. Incorporating creep into topology optimization (TO) remains open because the response is nonlinear, history-dependent, and thermomechanically coupled, and prior work often relies on linear viscoelastic models, which do not capture the behavior of metals at high temperatures. To bridge this gap, we introduce a differentiable thermo-structural TO framework. The approach considers creep deformation using the Norton model and leverages JAX's automatic differentiation to perform adjoint sensitivity analysis, enabling efficient gradient-based optimization. The transient material response is solved via a backward Euler scheme over a prescribed service life. Our objective is to minimize creep deformation subject to a volume constraint. We first demonstrate the framework on canonical two-dimensional benchmarks, showing that the proposed formulation significantly reduces permanent deformation compared to designs optimized solely for elastic stiffness. We then pose, as a challenge problem, the compositional design of a three-dimensional graded material turbine blade in which the local mixture of two candidate alloys is optimized. This challenge problem exercises the full capability of the framework, including transient nonlinear creep, coupled thermal loading, three-dimensional geometry, and gradient-based multi-material design, highlighting the need for creep-aware design in high-temperature applications.

cs.CE↗

arXiv2026

Towards Disentangled Preference Optimization Dynamics: Suppress the Loser, Preserve the Winner

Preference optimization is widely used to align large language models (LLMs) with human preferences. However, many margin-based methods also suppress the chosen response when they try to suppress the rejected one, and there is no general way to prevent this across different objectives. We address this issue with a unified incentive-score decomposition of preference optimization, revealing that different objectives share the same local update directions and differ only in their scalar weights. This decomposition provides a common framework for analyzing objectives that were previously studied in separate settings. Building on this decomposition, by analyzing the dynamics of the chosen/rejected likelihoods, we identify the disentanglement band (DB), a simple, testable condition that tells us when training can follow the desired path: suppress the loser while preserving the winner, possibly after an early stage. Using the DB, we propose reward calibration (RC), a plug-and-play method that adaptively rebalances the updates for chosen and rejected responses to satisfy the DB, without redesigning the base objective. Empirical results show that RC leads to more disentangled dynamics, with better downstream performance observed across several settings. Our code is available at https://github.com/IceyWuu/DisentangledPreferenceOptimization.

cs.LG↗