Search arXiv⌕ Search

arXiv subjects

Yue Wu

Publications and source records attributed to Yue Wu.

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

MorphoQuant: Modality-Aware Quantization for Omni-modal Large Language Models

Conventional Post-Training Quantization (PTQ) methods struggle with 4-bit Omni-modal Large Language Models (OLLMs) due to the extreme distribution heterogeneity and disparate outlier patterns across modalities. To address this, we propose MorphoQuant, a modality-aware PTQ framework engineered to preserve cross-modal morphology and mitigate outlier loss. Specifically, we introduce Distribution-Aware Bias Compensation (DABC), which selectively absorbs long-tailed outliers into channel-wise biases. This mechanism safeguards outlier magnitudes while maintaining high-precision discretization for dense inliers, thereby preserving accurate discretization across diverse modal distribution. Complementing this, we propose Morphology-Directed Quantization Function Optimization (MDQFO) to co-optimize the quantization grid with the bias mask, ensuring fine-grained alignment across modalities. Extensive evaluations on Qwen2.5-Omni across benchmarks like MMMU and Video-MME demonstrate our approach's superiority. Notably, our W4A4 model achieves 76.63% on ScienceQA, significantly outperforming SOTA W4A4 methods and surprisingly surpassing the W4A16 baseline, which fully demonstrates the exceptional accuracy-efficiency trade-off of our framework.

cs.CV↗
arXiv2026

Unraveling the Hidden Dynamical Structure in Recurrent Neural Policies

Recurrent neural policies are widely used in partially observable control and meta-RL tasks. Their abilities to maintain internal memory and adapt quickly to unseen scenarios have offered them unparalleled performance when compared to non-recurrent counterparts. However, until today, the underlying mechanisms for their superior generalization and robustness performance remain poorly understood. In this study, by analyzing the hidden state domain of recurrent policies learned over a diverse set of training methods, model architectures, and tasks, we find that stable cyclic structures consistently emerge during interaction with the environment. Such cyclic structures share a remarkable similarity with \textit{limit cycles} in dynamical system analysis, if we consider the policy and the environment as a joint hybrid dynamical system. Moreover, we uncover that the geometry of such limit cycles also has a structured correspondence with the policies' behaviors. These findings offer new perspectives to explain many nice properties of recurrent policies: the emergence of limit cycles stabilizes both the policies' internal memory and the task-relevant environmental states, while suppressing nuisance variability arising from environmental uncertainty; the geometry of limit cycles also encodes relational structures of behaviors, facilitating easier skill adaptation when facing non-stationary environments.

cs.LG↗
arXiv2026

Baton: Explicit Semantic Blueprints for Joint Video-Audio Generation

Current open-source diffusion models struggle to generate stable and synchronized audio-visual content, particularly in scenarios demanding complex semantic reasoning. The root cause is that existing methods rely on coarse text embeddings from off-the-shelf encoders to guide audio-video denoising, which discards fine-grained semantics and, critically, lacks a shared long-horizon plan, leading to uncoordinated denoising trajectories and fragile cross-modal alignment. We propose Baton, the first framework that introduces explicit semantic planning into joint video-audio generation. Our key insight is that complementing coarse text guidance with semantically rich, modality-aware planned tokens, jointly reasoned and mutually aligned before denoising, can simultaneously restore fine-grained semantic detail and establish a shared blueprint that coordinates both audio and video denoising trajectories. Concretely, Baton first introduces the VA-Planner, a multimodal language model equipped with dual semantic alignment towers, where learnable queries cross-attend to both video and audio features to produce a pair of semantically aligned video and audio planned tokens as keyframe-level blueprints. These planned tokens are injected into the diffusion backbone via cross-attention layers, providing temporally grounded guidance complementary to coarse text embeddings. Since planned tokens do not share one-to-one spatial-temporal correspondence with diffusion latents, we further propose Relative Semantic RoPE, a relative positional encoding that maps planned tokens and latents into a shared spatial-temporal coordinate frame, enabling each latent to accurately attend to its positionally corresponding semantic cues. Experiments on benchmarks show the effectiveness of Baton both qualitatively and quantitatively.

cs.CV↗
arXiv2026

Accelerating Min-Max Optimization via Power-Law Stepsizes

We revisit the convergence guarantees of the Extragradient (EG) method for unconstrained biaffine min-max optimization. It is known that EG with a fixed stepsize achieves a $Θ(T^{-1/2})$ last-iterate convergence rate, which is slower than the optimal $\mathcal{O}(T^{-1})$ rate attainable by incorporating additional mechanisms such as anchoring. Motivated by recent advances showing that dynamic stepsizes alone can significantly accelerate gradient descent, we ask whether dynamic stepsizes can similarly accelerate the last-iterate convergence of EG. We present the first positive result in this direction. Specifically, we provide a deterministic dynamic stepsize schedule that accelerates the convergence rate of EG to $\mathcal{O}(T^{-2/3+\varepsilon})$ for any $\varepsilon > 0$. We also show that this rate is tight when the extrapolation and update steps of EG use the same stepsize. We then show that allowing different stepsizes for the extrapolation and update steps further improves the convergence rate to the near-optimal $\mathcal{O}(T^{-1+\varepsilon})$. Our analysis reduces stepsize scheduling to an optimization problem, whose solution leads to a stepsize schedule that follows (a discretization of) a power-law distribution. Our proposed stepsize schedules and analysis extend to other methods, such as Optimistic Gradient (OG), and suggest broader applicability to general min-max optimization problems.

math.OC↗
arXiv2026

Bending-like stress induced by solder joint under uniaxial tensile testing in 2G-HTS tapes: Impact and optimization approach

The reversible stress limit (\mathit{R}_{rev}) of second-generation high-temperature superconducting (2G-HTS) tapes is a critical performance indicator, typically characterized through uniaxial tensile testing. In practice, the accuracy of the measured Rrev value is often compromised by stress concentration induced by the voltage tap solder joint. The present study investigates the underlying interference mechanism using integrated experimental and numerical methods. Mechanistic analysis reveals that under uniaxial tensile loading, the local geometric inhomogeneity introduced by the solder joint induces an external, bending-like stress in the vicinity of the solder joint, transitioning from additional tensile stress in the zone adjacent to the joint to additional compressive stress in the zone remote from it. When the solder joint is attached to the front surface of the tape (the side closer to the superconducting layer), the superconducting layer experiences localized additional tensile stress, triggering premature damage and early \mathit{I}_{c} degradation. Consequently, an optimized back-surface soldering approach is proposed, which positions the superconducting layer in a localized compressive zone. Experimental validation demonstrates that the proposed approach effectively mitigates testing errors for various tape configurations. Notably, for the tape with a copper layer thickness of 5 μm, the measured \mathit{R}_{rev} increased from 546 MPa to 734 MPa, corresponding to a 42% increase, and moved closer to the actual value. The findings provide essential insights for the precision characterization of the electromechanical performances (EMPs) of 2G-HTS tapes.

cond-mat.supr-con↗
arXiv2026

A Parameter-Free Restart Scheme with Only a Parallelizable $\log\log(1/ε)$ Overhead

It is well-known that first-order methods can offer accelerated convergence rates in the presence of growth structures. Restarting schemes provide a general tool for such speed-ups. These schemes typically either require unrealistic problem knowledge, incur logarithmic overhead factors in oracle complexity, and/or have a nontrivial initial burn-in phase. We present a parameter-free approach for restarting any first-order method, avoiding these three drawbacks. Our approach dynamically deploys parallel instances of a given first-order method communicating progress in the style of Renegar and Grimmer. Our optimized scheme avoids expensive burn-ins and only requires $O(\log\log(1/ε))$ parallel processes when the accelerated rate is sublinear.

math.OC↗
arXiv2026

Precise: SDE-Consistent Stochastic Sampling for RL Post-Training of Flow-Matching Models

Reinforcement learning (RL) has become an effective way to improve prompt alignment and perceptual quality in diffusion and flow-matching generators. A critical step for applying online RL to flow matching is turning the deterministic sampling trajectory into a stochastic policy, typically by replacing the reverse-time Ordinary Differential Equation (ODE) with a Stochastic Differential Equation (SDE). The stochastic sampler, controlling the exploration behavior and denoising dynamics, is thus part of the policy, and its design can significantly affect the reward optimization performance. We break down the sampler design into two interdependent components: choosing the right amount of stochastic exploration, and discretizing the resulting SDE faithfully at the small step counts used in RL. To address the first component, we analyze the inherent tension between exploration and stability in denoising and derive an SDE schedule that balances the two. Turning to the discretization challenge, we use a toy example to show that existing samplers can deviate from the flow-matching process, either by introducing excessive discretization noise or by relying on heuristic rules that do not guarantee convergence to the data distribution. To address these issues, we propose Precise, a new stochastic sampler that balances effective exploration with stability. Crucially, Precise keeps the denoising trajectory SDE-consistent through a novel approximation that freezes the clean-latent posterior mean, resolving the excess noise issue in standard samplers. Extensive experiments demonstrate that this formulation leads to significantly faster and more stable reward optimization via reinforcement learning, achieving state-of-the-art alignment scores (e.g., PickScore, HPSv2.1) while requiring 13.1-53.2% less wall-clock training time to match the best in-domain performance of prior samplers.

cs.LG↗
arXiv2026

Accelerating 3D Gaussian Splatting using Tensor Cores

3D Gaussian Splatting (3DGS) has become a leading technique for real-time neural rendering and 3D scene reconstruction, but its rendering cost remains too high for many latency-sensitive scenarios. In particular, the rasterization stage in 3DGS dominates end-to-end rendering time, during which the renderer repeatedly evaluates each Gaussian's contribution to each covered pixel, making this stage compute-bound. At the same time, modern GPUs provide high-throughput Tensor Cores for low-precision matrix operations, yet existing 3DGS systems execute rasterization entirely on CUDA cores and leave Tensor Cores idle. We find that 3DGS rendering can be executed in FP16 with negligible quality degradation, suggesting a promising opportunity for Tensor Core acceleration. However, exploiting Tensor Cores for 3DGS is non-trivial because rasterization does not naturally match their execution model. Existing 3DGS rasterization is expressed as irregular per-pixel scalar operations, whereas Tensor Cores require dense, regular, and reuse-rich matrix workloads. Moreover, conventional tile-by-tile execution fails to exploit Gaussian reuse across neighboring tiles, resulting in repeated data loading and thus high data movement overhead. To this end, we present TensorGS, a 3DGS acceleration framework using Tensor Cores. TensorGS tensorizes the dominant rasterization computation into Tensor-Core-compatible matrix operations and introduces cross-tile grouping to improve Gaussian reuse, amortize overhead, and increase Tensor Core utilization. Experimental results show that TensorGS improves end-to-end rendering performance by 1.65$\times$ while preserving image quality.

cs.GR↗
arXiv2026

Control-Channel Informativity for Koopman EDMDc under Behavior-Policy Data

Extended dynamic mode decomposition with control (EDMDc) is often trained from trajectories generated by a behavior policy or a pre-existing feedback controller. Such data can predict the observed behavior accurately while failing to identify how new input commands change the lifted state. This paper studies that failure as a control-channel informativity problem. We introduce a conditional intervention certificate, defined as the residual input covariance after projecting the input data away from the active lifted-state feature span. The certificate is the Schur complement of the lifted-state block in the EDMDc information matrix. We prove that its strict positivity is necessary and sufficient for finite-sample sample- identifiability of the lifted control-channel block. If the certificate vanishes, distinct lifted models agree on every collected transition but disagree under counterfactual inputs. We then give a closed-loop statistical bound using predictable regressors, conditionally sub-Gaussian transition noise, and a regularized Schur complement. A scalar feedback example shows the unavoidable scaling: under dithered feedback, residual intervention information grows quadratically with the dither amplitude and the control-channel error decreases with the inverse intervention signal-to-noise scale. New experiments verify these scalings exactly in a linear system and diagnostically in controlled Duffing and Van der Pol benchmarks. A larger EDMDc acquisition grid further shows that state coverage, joint regression conditioning, and intervention excitation are complementary diagnostics rather than interchangeable performance score.

math.OC↗
arXiv2026

Diagnostic Certificates of Data Quality and Regression Identifiability for Koopman Identification

Classical persistent excitation criteria usually assess whether an input or regressor signal is sufficiently rich. In Koopman and EDMD with control (EDMDc), however, data quality is determined by the concatenation of lifted state features and control inputs. Input-rich data can still visit a narrow state region, well-spread state samples can still produce degenerate lifted features, and both can fail to condition the final regression problem. This paper develops a diagnostic certificate framework for locating these failures. The certificates separate state-space coverage and clustering, lifted-feature nondegeneracy, and the final regression spectrum. The regression-spectrum certificate is the layer with direct theoretical guarantees: it controls the active standardized design's smallest singular value, has Fisher-information and one-step EDMDc stability interpretations, and admits a finite-sample lower bound under a population spectral gap. We also give structural examples and a Schur-complement condition showing why state, lifted, input, and regression diagnostics cannot be substituted for one another. As a sampling example, IGPE-DOPT uses these certificates to score candidate trajectory segments. Experiments on Duffing, Van der Pol, and Lorenz systems compare input-, state-, lifted-, and regression-oriented baselines. The results show that certificate layers separate, budget and weights shift bottlenecks, and downstream prediction or control performance is not monotone in any single certificate. The framework is therefore intended as an interpretable diagnostic and data-collection guide, not as a universal optimality claim.

math.OC↗
arXiv2026

Physics-informed operator learning for transferable energy-dissipative microstructure dynamics

Phase-field simulations provide mechanistic descriptions of microstructure evolution, but repeated high-fidelity integration over long horizons and broad parameter spaces remains computationally expensive. We present PFNet, a physics-informed neural operator framework that advances microstructural states by learning conditional evolution operators rather than direct correlations. PFNet combines a diffusion-inspired U-Net with periodic padding, entropy-based state conditioning and thermodynamic-parameter modulation to encode boundary consistency, instantaneous ordering state and changes in the free-energy landscape. For Cahn-Hilliard coarsening, PFNet achieves accurate one-step prediction and stable autoregressive rollouts across composition, gradient-energy coefficient, coarsening stage and morphology class, with errors concentrated near diffuse interfaces and topology-changing regions. The same framework extends to a four-channel martensitic-transformation benchmark without martensite-specific redesign. These results indicate that physics-informed operator learning can provide transferable surrogates for phase-field dynamics and broader energy-dissipative dynamical systems.

cond-mat.mtrl-sci↗
arXiv2026

AcademiClaw: When Students Set Challenges for AI Agents

Benchmarks within the OpenClaw ecosystem have thus far evaluated exclusively assistant-level tasks, leaving the academic-level capabilities of OpenClaw largely unexamined. We introduce AcademiClaw, a bilingual benchmark of 80 complex, long-horizon tasks sourced directly from university students' real academic workflows -- homework, research projects, competitions, and personal projects -- that they found current AI agents unable to solve effectively. Curated from 230 student-submitted candidates through rigorous expert review, the final task set spans 25+ professional domains, ranging from olympiad-level mathematics and linguistics problems to GPU-intensive reinforcement learning and full-stack system debugging, with 16 tasks requiring CUDA GPU execution. Each task executes in an isolated Docker sandbox and is scored on task completion by multi-dimensional rubrics combining six complementary techniques, with an independent five-category safety audit providing additional behavioral analysis. Experiments on six frontier models show that even the best achieves only a 55\% pass rate. Further analysis uncovers sharp capability boundaries across task domains, divergent behavioral strategies among models, and a disconnect between token consumption and output quality, providing fine-grained diagnostic signals beyond what aggregate metrics reveal. We hope that AcademiClaw and its open-sourced data and code can serve as a useful resource for the OpenClaw community, driving progress toward agents that are more capable and versatile across the full breadth of real-world academic demands. All data and code are available at https://github.com/GAIR-NLP/AcademiClaw.

cs.AI↗
arXiv2026

SDLLMFuzz: Dynamic-static LLM-assisted greybox fuzzing for structured input programs

Fuzzing has become a widely adopted technique for vulnerability discovery, yet it remains ineffective for structured-input programs due to strict syntactic constraints and limited semantic awareness. Traditional greybox fuzzers rely on mutation-based strategies and coarse-grained coverage feedback, which often fail to generate valid inputs and explore deep execution paths. Recent advances in large language models (LLMs) have shown promise in improving input generation, but existing approaches primarily focus on seed generation and largely overlook the effective use of runtime feedback. In this paper, we propose SDLLMFuzz, a dynamic-static LLM-assisted greybox fuzzing framework for structured-input programs. Our approach integrates LLM-based structure-aware seed generation with static crash analysis, forming a unified feedback loop that iteratively refines test inputs. Specifically, we leverage LLMs to generate syntactically valid and semantically diverse inputs, while extracting rich semantic information from crash artifacts (e.g., core dumps and execution traces) to guide subsequent input generation. This dynamic-static feedback mechanism enables more efficient exploration of complex program behaviors. We evaluate SDLLMFuzz on the Magma benchmark across multiple structured-input programs, including libxml2, libpng, and libsndfile. Experimental results show that SDLLMFuzz significantly outperforms traditional greybox fuzzers and LLM-assisted baselines in terms of bug discovery and time-to-bug. These results demonstrate that combining semantic input generation with feedback-driven refinement is an effective direction for improving fuzzing performance on structured-input programs.

cs.CR↗
arXiv2026

Stability-Weighted Decoding for Diffusion Language Models

Diffusion large language models (dLLMs) enable parallel text generation by iteratively denoising a fully masked sequence, unmasking a subset of masked tokens at each step. Existing decoding strategies rely on static confidence metrics computed at a single denoising step, ignoring temporal history and often leading to premature unmasking of unstable tokens. In this work, we theoretically establish that a token's temporal instability, quantified by the KL divergence between consecutive prediction distributions, provides a strict lower bound on its mutual information with the remaining masked context, indicating that temporally unstable tokens are inherently unsafe to unmask. Based on this insight, we propose Stability-Weighted Decoding (SWD), a training-free, plug-and-play strategy that incorporates temporal stability into token scoring and acts as a universal modulator for arbitrary score-based decoding policies. Experiments on code generation and mathematical reasoning benchmarks demonstrate that SWD consistently improves generation accuracy across representative scoring metrics and selection policies, and exhibits exceptional robustness, maintaining a significant performance lead over standard baselines across varying acceleration ratios.

cs.CL↗
arXiv2026

FAST: A Synergistic Framework of Attention and State-space Models for Spatiotemporal Traffic Prediction

Traffic forecasting requires modeling complex temporal dynamics and long-range spatial dependencies over large sensor networks. Existing methods typically face a trade-off between expressiveness and efficiency: Transformer-based models capture global dependencies well but suffer from quadratic complexity, while recent selective state-space models are computationally efficient yet less effective at modeling spatial interactions in graph-structured traffic data. We propose FAST, a unified framework that combines attention and state-space modeling for scalable spatiotemporal traffic forecasting. FAST adopts a Temporal-Spatial-Temporal architecture, where temporal attention modules capture both short- and long-term temporal patterns, and a Mamba-based spatial module models long-range inter-sensor dependencies with linear complexity. To better represent heterogeneous traffic contexts, FAST further introduces a learnable multi-source spatiotemporal embedding that integrates historical traffic flow, temporal context, and node-level information, together with a multi-level skip prediction mechanism for hierarchical feature fusion. Experiments on PeMS04, PeMS07, and PeMS08 show that FAST consistently outperforms strong baselines from Transformer-, GNN-, attention-, and Mamba-based families. In particular, FAST achieves the best MAE and RMSE on all three benchmarks, with up to 4.3\% lower RMSE and 2.8\% lower MAE than the strongest baseline, demonstrating a favorable balance between accuracy, scalability, and generalization.

cs.LG↗
arXiv2026

GeM-EA: A Generative and Meta-learning Enhanced Evolutionary Algorithm for Streaming Data-Driven Optimization

Streaming Data-Driven Optimization (SDDO) problems arise in many applications where data arrive continuously and the optimization environment evolves over time. Concept drift produces non-stationary landscapes, making optimization methods challenging due to outdated models. Existing approaches often rely on simple surrogate combinations or directly injecting solutions, which may cause negative transfer under sudden environmental changes. We propose GeM-EA, a Generative and Meta-learning Enhanced Evolutionary Algorithm for SDDO that unifies meta-learned surrogate adaptation with generative replay for effective evolutionary search. Upon detecting concept drift, a bi-level meta-learning strategy rapidly initializes the surrogate using environment-relevant priors, while a linear residual component captures global trends. A multi-island evolutionary strategy further leverages historical knowledge via generative replay to accelerate optimization. Experimental results on benchmark SDDO problems demonstrate that GeM-EA achieves faster adaptation and improved robustness compared with state-of-the-art methods.

cs.NE↗
arXiv2026

Learning to Focus and Precise Cropping: A Reinforcement Learning Framework with Information Gaps and Grounding Loss for MLLMs

To enhance the perception and reasoning capabilities of multimodal large language models in complex visual scenes, recent research has introduced agent-based workflows. In these works, MLLMs autonomously utilize image cropping tool to analyze regions of interest for question answering. While existing training strategies, such as those employing supervised fine-tuning and reinforcement learning, have made significant progress, our empirical analysis reveals a key limitation. We demonstrate the model's strong reliance on global input and its weak dependence on the details within the cropped region. To address this issue, we propose a novel two-stage reinforcement learning framework that does not require trajectory supervision. In the first stage, we introduce the ``Information Gap" mechanism by adjusting the granularity of the global image. This mechanism trains the model to answer questions by focusing on cropped key regions, driven by the information gain these regions provide. The second stage further enhances cropping precision by incorporating a grounding loss, using a small number of bounding box annotations. Experiments show that our method significantly enhances the model's attention to cropped regions, enabling it to achieve state-of-the-art performance on high-resolution visual question-answering benchmarks. Our method provides a more efficient approach for perceiving and reasoning fine-grained details in MLLMs. Code is available at: https://github.com/XuanPu-Z/LFPC.

cs.CV↗
arXiv2026

Exceptionally Slow Relaxation from Micro-canonical to Canonical Ensembles in Quasi-one-dimensional Quantum Gases

Integrability in one dimension prevents quantum thermalization and gives rise to rich many-body phenomena described by generalized hydrodynamics, which have been extensively studied over the past two decades using cold atoms in optically confined tubes. However, experimental work to date has focused primarily on low-energy states. Here, we report the experimental observation and theoretical understanding of near-integrable effects on thermalization in highly excited states. We design a protocol to prepare atoms within a high-energy window by combining a harmonic trap and a weak optical lattice: a Bose-Einstein condensate is initially prepared away from the trap center via Wannier-Stark localization and subsequently emits atoms into a selected energy window of highly excited states via Landau-Zener tunneling. By reconstructing the Wigner functions from the density distribution using a machine learning algorithm, we find that it takes an exceptionally long time, up to several seconds, for these atoms to gradually thermalize from an approximately microcanonical ensemble toward a canonical ensemble. We develop a modified Boltzmann equation that captures weak integrability breaking, yielding good agreement between theory and experiment. Our results extend the understanding of integrability and thermalization in low-dimensional quantum systems.

cond-mat.quant-gas↗