arXiv subjects

Yong Wu

Publications and source records attributed to Yong Wu.

At least 37 records · Page 2Linked to original sources

arXiv2026

Multiple charge transfer driven complex reaction dynamics: covalent bonding meets van der Waals interactions

Ultrafast charge transfer (CT) processes redistribute electronic charge within and between molecular units and play a central role in many physical, chemical, and biological phenomena. However, the microscopic pathways of multiple CT events, including the coupled structural evolution and energy redistribution, are challenging to disentangle experimentally in complex systems. To obtain controlled insight into such dynamics, well-defined properties are required. Here, we investigate the N2Ar dimer, which combines a covalent bond with a weak van der Waals interaction, using site-selective synchrotron photoionization and coincident detection of electrons and ions. Combined with ab initio calculations, this approach enables step-by-step tracking of ultrafast CT and fragmentation dynamics. We find that the dimer's structural evolution triggers a second CT event, opening complex reaction pathways in which electrons are transferred back and forth between Ar and N2, through two nonadiabatic transitions involving conical intersections. These results demonstrate that sequential multiple CT-induced transitions, even in a simple dimer, provide controlled insight into nonadiabatic reaction mechanisms relevant to complex systems.

physics.chem-ph↗

arXiv2026

FlashInfer-Bench: Building the Virtuous Cycle for AI-driven LLM Systems

Recent advances show that large language models (LLMs) can act as autonomous agents capable of generating GPU kernels, but integrating these AI-generated kernels into real-world inference systems remains challenging. FlashInfer-Bench addresses this gap by establishing a standardized, closed-loop framework that connects kernel generation, benchmarking, and deployment. At its core, FlashInfer Trace provides a unified schema describing kernel definitions, workloads, implementations, and evaluations, enabling consistent communication between agents and systems. Built on real serving traces, FlashInfer-Bench includes a curated dataset, a robust correctness- and performance-aware benchmarking framework, a public leaderboard to track LLM agents' GPU programming capabilities, and a dynamic substitution mechanism (apply()) that seamlessly injects the best-performing kernels into production LLM engines such as SGLang and vLLM. Using FlashInfer-Bench, we further evaluate the performance and limitations of LLM agents, compare the trade-offs among different GPU programming languages, and provide insights for future agent design. FlashInfer-Bench thus establishes a practical, reproducible pathway for continuously improving AI-generated kernels and deploying them into large-scale LLM inference.

cs.AI↗

arXiv2025

Index-ASR Technical Report

Automatic speech recognition (ASR) has witnessed remarkable progress in recent years, largely driven by the emergence of LLM-based ASR paradigm. Despite their strong performance on a variety of open-source benchmarks, existing LLM-based ASR systems still suffer from two critical limitations. First, they are prone to hallucination errors, often generating excessively long and repetitive outputs that are not well grounded in the acoustic input. Second, they provide limited support for flexible and fine-grained contextual customization. To address these challenges, we propose Index-ASR, a large-scale LLM-based ASR system designed to simultaneously enhance robustness and support customizable hotword recognition. The core idea of Index-ASR lies in the integration of LLM and large-scale training data enriched with background noise and contextual information. Experimental results show that our Index-ASR achieves strong performance on both open-source benchmarks and in-house test sets, highlighting its robustness and practicality for real-world ASR applications.

cs.SD↗

arXiv2025

Uncertainty-Preserving QBNNs: Multi-Level Quantization of SVI-Based Bayesian Neural Networks for Image Classification

Bayesian Neural Networks (BNNs) provide principled uncertainty quantification but suffer from substantial computational and memory overhead compared to deterministic networks. While quantization techniques have successfully reduced resource requirements in standard deep learning models, their application to probabilistic models remains largely unexplored. We introduce a systematic multi-level quantization framework for Stochastic Variational Inference based BNNs that distinguishes between three quantization strategies: Variational Parameter Quantization (VPQ), Sampled Parameter Quantization (SPQ), and Joint Quantization (JQ). Our logarithmic quantization for variance parameters, and specialized activation functions to preserve the distributional structure are essential for calibrated uncertainty estimation. Through comprehensive experiments on Dirty-MNIST, we demonstrate that BNNs can be quantized down to 4-bit precision while maintaining both classification accuracy and uncertainty disentanglement. At 4 bits, Joint Quantization achieves up to 8x memory reduction compared to floating-point implementations with minimal degradation in epistemic and aleatoric uncertainty estimation. These results enable deployment of BNNs on resource-constrained edge devices and provide design guidelines for future analog "Bayesian Machines" operating at inherently low precision.

cs.LG↗

arXiv2025

P-orbital spin generator with large spin Hall angle and long spin diffusion length

High density data storage and spin-logic devices require highly efficient all-electric control of spin moments. So far, charge-to-spin conversion through the spin Hall effect (SHE) highly limits to d-orbital materials associated with strong spin-orbit coupling (SOC), especially heavy metals. However, d-orbital heavy metals with strong SOC results in a short spin diffusion length, which restricts the spin transport and accumulation in spintronic devices. Therefore, it is urgent to discovery new SHE materials with both large spin Hall conductivity and high spin transport ability beyond d-orbital materials. Here, we experimentally report a large charge to spin conversion in a p-orbital In2Bi alloy, exhibiting the coexistence of a large spin Hall angle and a long spin diffusion length (4 times that of Pt). Our first-principles calculations reveal that small gap openings near the Fermi level lead to large Berry curvature-related spin Hall conductivity. Due to the delocalized nature of p-orbitals of In2Bi, its spin current can overcome the physical barriers between spin Hall angle and spin diffusion length in d-orbital metals, thereby advancing the development of high performance spintronic devices.

cond-mat.mtrl-sci↗

arXiv2025

Discovering Causal Relationships using Proxy Variables under Unmeasured Confounding

Inferring causal relationships between variable pairs in the observational study is crucial but challenging, due to the presence of unmeasured confounding. While previous methods employed the negative controls to adjust for the confounding bias, they were either restricted to the discrete setting (i.e., all variables are discrete) or relied on strong assumptions for identification. To address these problems, we develop a general nonparametric approach that accommodates both discrete and continuous settings for testing causal hypothesis under unmeasured confounders. By using only a single negative control outcome (NCO), we establish a new identification result based on a newly proposed integral equation that links the outcome and NCO, requiring only the completeness and mild regularity conditions. We then propose a kernel-based testing procedure that is more efficient than existing moment-restriction methods. We derive the asymptotic level and power properties for our tests. Furthermore, we examine cases where our procedure using only NCO fails to achieve identification, and introduce a new procedure that incorporates a negative control exposure (NCE) to restore identifiability. We demonstrate the effectiveness of our approach through extensive simulations and real-world data from the Intensive Care Data and World Values Survey.

stat.ME↗

arXiv2025

PEVLM: Parallel Encoding for Vision-Language Models

Vision-Language Models (VLMs) have demonstrated strong capabilities in multimodal understanding and generation tasks. However, their application to long video understanding remains hindered by the quadratic complexity of standard attention mechanisms. In this work, we introduce \textbf{PEVLM}, a fine-tuning-free parallel encoding method designed to enhance the prefilling efficiency of VLMs in long video scenarios. PEVLM partitions the input video into context blocks with a shared sink block, while preserving sequential position embeddings to align the attention weight distribution with that of Full-Attention. This design reduces attention complexity from $O((T \times N)^2)$ to $O(T \times N)$ where $T$ is the number of frames and $N$ the number of tokens per frame, without sacrificing accuracy. Extensive experiments across multiple state-of-the-art models and benchmarks demonstrate that PEVLM consistently outperforms existing parallel encoding approaches, achieving up to \textbf{7.47x} speedup in attention computation and reducing end-to-end latency by \textbf{40\%}. Remarkably, PEVLM not only maintains high accuracy, but in some settings even surpasses Full-Attention performance. Under strict latency constraints, it achieves substantial gains, improving accuracy from \textbf{23.26\%} to \textbf{61.03\%}. These results underscore the effectiveness of PEVLM for low-latency, long-context video understanding, making it a promising solution for real-world applications.

cs.CV↗

arXiv2025

Beyond Templates: Dynamic Adaptation of Reasoning Demonstrations via Feasibility-Aware Exploration

Large language models (LLMs) have shown remarkable reasoning capabilities, yet aligning such abilities to small language models (SLMs) remains a challenge due to distributional mismatches and limited model capacity. Existing reasoning datasets, typically designed for powerful LLMs, often lead to degraded performance when directly applied to weaker models. In this work, we introduce Dynamic Adaptation of Reasoning Trajectories (DART), a novel data adaptation framework that bridges the capability gap between expert reasoning trajectories and diverse SLMs. Instead of uniformly imitating expert steps, DART employs a selective imitation strategy guided by step-wise adaptability estimation via solution simulation. When expert steps surpass the student's capacity -- signaled by an Imitation Gap -- the student autonomously explores alternative reasoning paths, constrained by outcome consistency. We validate DART across multiple reasoning benchmarks and model scales, demonstrating that it significantly improves generalization and data efficiency over static fine-tuning. Our method enhances supervision quality by aligning training signals with the student's reasoning capabilities, offering a scalable solution for reasoning alignment in resource-constrained models.

cs.CL↗

arXiv2025

Balanced 3DGS: Gaussian-wise Parallelism Rendering with Fine-Grained Tiling

3D Gaussian Splatting (3DGS) is increasingly attracting attention in both academia and industry owing to its superior visual quality and rendering speed. However, training a 3DGS model remains a time-intensive task, especially in load imbalance scenarios where workload diversity among pixels and Gaussian spheres causes poor renderCUDA kernel performance. We introduce Balanced 3DGS, a Gaussian-wise parallelism rendering with fine-grained tiling approach in 3DGS training process, perfectly solving load-imbalance issues. First, we innovatively introduce the inter-block dynamic workload distribution technique to map workloads to Streaming Multiprocessor(SM) resources within a single GPU dynamically, which constitutes the foundation of load balancing. Second, we are the first to propose the Gaussian-wise parallel rendering technique to significantly reduce workload divergence inside a warp, which serves as a critical component in addressing load imbalance. Based on the above two methods, we further creatively put forward the fine-grained combined load balancing technique to uniformly distribute workload across all SMs, which boosts the forward renderCUDA kernel performance by up to 7.52x. Besides, we present a self-adaptive render kernel selection strategy during the 3DGS training process based on different load-balance situations, which effectively improves training efficiency.

cs.CV↗

arXiv2025

Cognitive Memory in Large Language Models

This paper examines memory mechanisms in Large Language Models (LLMs), emphasizing their importance for context-rich responses, reduced hallucinations, and improved efficiency. It categorizes memory into sensory, short-term, and long-term, with sensory memory corresponding to input prompts, short-term memory processing immediate context, and long-term memory implemented via external databases or structures. The text-based memory section covers acquisition (selection and summarization), management (updating, accessing, storing, and resolving conflicts), and utilization (full-text search, SQL queries, semantic search). The KV cache-based memory section discusses selection methods (regularity-based summarization, score-based approaches, special token embeddings) and compression techniques (low-rank compression, KV merging, multimodal compression), along with management strategies like offloading and shared attention mechanisms. Parameter-based memory methods (LoRA, TTT, MoE) transform memories into model parameters to enhance efficiency, while hidden-state-based memory approaches (chunk mechanisms, recurrent transformers, Mamba model) improve long-text processing by combining RNN hidden states with current methods. Overall, the paper offers a comprehensive analysis of LLM memory mechanisms, highlighting their significance and future research directions.

cs.CL↗

arXiv2025

Implementation of full-potential screened spherical wave based muffin-tin orbital for all-electron density functional theory

Screened spherical wave (SSW) of the Hankel function features the complete, minimal and short-ranged basis set, presenting a compact representation for electronic systems. In this work, we report the implementation of full-potential (FP) SSW based tight-binding linearized Muffin-Tin orbital (TB-LMTO) for all-electron density functional theory (DFT), and provide extensive tests on the robustness of FP-TB-LMTO and its high accuracy for first-principles material simulation. Through the introduction of double augmentation, SSW based MTO is accurately represented on the double grids including the full-space uniform and dense radial grids. Based on the the double augmentation, the accurate computation of full charge density, full potential,complex integral in the interstitial region and the total energy are all effectively addressed to realize the FP-TB-LMTO for DFT self-consistent calculations. By calculating the total energy,band structure, phase ordering, and elastic constants for a wide variety of materials, including normal metals, compounds, and diamond silicon, we domenstrate the highly accurate numerical implemetation of FP-TB-LMTO for all-electron DFT in comparison with other well-established FP method. The implementation of FP-TB-LMTO based DFT offers an important tool for the accurate first-principles tight-binding electronic structure calculations, particular important for the large-scale or strongly correlated materials.

cond-mat.mtrl-sci↗

arXiv2025

Relax: Composable Abstractions for End-to-End Dynamic Machine Learning

Dynamic shape computations have become critical in modern machine learning workloads, especially in emerging large language models. The success of these models has driven the demand for their universal deployment across a diverse set of backend environments. In this paper, we present Relax, a compiler abstraction for optimizing end-to-end dynamic machine learning workloads. Relax introduces a cross-level abstraction that encapsulates computational graphs, loop-level tensor programs, and external library calls in a single representation. Relax also introduces first-class symbolic shape annotations to track dynamic shape computations globally across the program, enabling dynamic shape-aware cross-level optimizations. We build an end-to-end compilation framework using the proposed approach to optimize dynamic shape models. Experimental results on LLMs show that Relax delivers performance competitive with state-of-the-art systems across various GPUs and enables deployment of emerging models to a broader set of emerging environments, including mobile phones, embedded devices, and web browsers.

cs.LG↗

arXiv2024

Dynamic learning of synchronization in coupled nonlinear systems

Synchronization phenomena are pervasive in coupled nonlinear systems across the natural world and engineering domains. Understanding how to dynamically identify the parameter space (or network structure) of coupled nonlinear systems in a synchronized state is crucial for the study of system synchronization. To address the challenge of achieving stable synchronization in coupled nonlinear systems, we develop a set of mathematical optimization techniques for dynamic learning of synchronization (DLS) inspired by machine learning. This technology captures the state differences between nodes within the system and dynamically adjusts weights, allowing coupled nonlinear systems to maintain a stable state of synchronization after appropriate weight adjustments. To enhance synchronization optimization, we use the Master Stability Function (MSF) to demonstrate how DLS effectively adjusts networks into their synchronization regions. We introduce several variants of the DLS technique, including adaptive, supervised, and hybrid methods, effectively promoting synchronization in heterogeneous networks such as small-world, scale-free, and random networks. The efficacy of this technique is validated through its application to simple FitzHugh-Nagumo neural networks and complex Hodgkin-Huxley neuronal networks, examining its impact on both global and local synchronization. The DLS technique proposed in this study offers a new solution to synchronization problems in dynamic network environments, addressing the deficiencies in adaptability and flexibility of existing technologies and providing a fresh perspective for understanding and implementing synchronization phenomena in coupled nonlinear systems.

physics.bio-ph↗

arXiv2024

Generation of cold polyatomic cations by cascade reactive two-body ion-atom collisions

Polyatomic cations $^{87}$Rb$_M^+$ ($M$ = 2, 3,$\ldots$) have been produced by cascade two-body ion-atom reactive collisions in the two-step CW-laser photoionization of laser-cooled $^{87}$Rb atoms and accumulated in the ion trap. Using resonant-excitation mass spectrometry and resonant excitation-assisted time-of-flight mass spectrometry, we directly observed and distinguished the charged reaction products. We experimentally verified the cascade generation and cascade dissociation of $^{87}$Rb$_M^+$. The populations of $^{87}$Rb$_M^+$ are quantitatively investigated by solving the rate equations. The $^{87}$Rb$^+$-$^{87}$Rb reaction rate coefficient was derived as 9.10$\times10^{-11}$ cm$^3$/s accordingly. The methods developed here for assembling and detecting homonuclear polyatomic cations can be applied to any experiment in ion-atom hybrid traps. The present study lays the foundation for exploring atomically precise metal clusters and physics from few- to many-body perspective.

physics.atom-ph↗

arXiv2024

ELF-UA: Efficient Label-Free User Adaptation in Gaze Estimation

We consider the problem of user-adaptive 3D gaze estimation. The performance of person-independent gaze estimation is limited due to interpersonal anatomical differences. Our goal is to provide a personalized gaze estimation model specifically adapted to a target user. Previous work on user-adaptive gaze estimation requires some labeled images of the target person data to fine-tune the model at test time. However, this can be unrealistic in real-world applications, since it is cumbersome for an end-user to provide labeled images. In addition, previous work requires the training data to have both gaze labels and person IDs. This data requirement makes it infeasible to use some of the available data. To tackle these challenges, this paper proposes a new problem called efficient label-free user adaptation in gaze estimation. Our model only needs a few unlabeled images of a target user for the model adaptation. During offline training, we have some labeled source data without person IDs and some unlabeled person-specific data. Our proposed method uses a meta-learning approach to learn how to adapt to a new user with only a few unlabeled images. Our key technical innovation is to use a generalization bound from domain adaptation to define the loss function in meta-learning, so that our method can effectively make use of both the labeled source data and the unlabeled person-specific data during training. Extensive experiments validate the effectiveness of our method on several challenging benchmarks.

cs.CV↗

arXiv2024

Error estimates of a regularized finite difference method for the Logarithmic Schrödinger equation with Dirac delta potential

In this paper, we introduce a conservative Crank-Nicolson-type finite difference schemes for the regularized logarithmic Schrödinger equation (RLSE) with Dirac delta potential in 1D. The regularized logarithmic Schrödinger equation with a small regularized parameter $0<\eps \ll 1$ is adopted to approximate the logarithmic Schrödinger equation (LSE) with linear convergence rate $O(\eps)$. The numerical method can be used to avoid numerical blow-up and/or to suppress round-off error due to the logarithmic nonlinearity in LSE. Then, by using domain-decomposition technique, we can transform the original problem into an interface problem. Different treatments on the interface conditions lead to different discrete schemes and it turns out that a simple discrete approximation of the Dirac potential coincides with one of the conservative finite difference schemes. The optimal $H^1$ error estimates and the conservative properties of the finite difference schemes are investigated. The Crank-Nicolson finite difference methods enjoy the second-order convergence rate in time and space. Numerical examples are provided to support our analysis and show the accuracy and efficiency of the numerical method.

math.NA↗

arXiv2024

NTIRE 2024 Challenge on Short-form UGC Video Quality Assessment: Methods and Results

This paper reviews the NTIRE 2024 Challenge on Shortform UGC Video Quality Assessment (S-UGC VQA), where various excellent solutions are submitted and evaluated on the collected dataset KVQ from popular short-form video platform, i.e., Kuaishou/Kwai Platform. The KVQ database is divided into three parts, including 2926 videos for training, 420 videos for validation, and 854 videos for testing. The purpose is to build new benchmarks and advance the development of S-UGC VQA. The competition had 200 participants and 13 teams submitted valid solutions for the final testing phase. The proposed solutions achieved state-of-the-art performances for S-UGC VQA. The project can be found at https://github.com/lixinustc/KVQChallenge-CVPR-NTIRE2024.

eess.IV↗

arXiv2024

Doubly Robust Proximal Causal Learning for Continuous Treatments

Proximal causal learning is a promising framework for identifying the causal effect under the existence of unmeasured confounders. Within this framework, the doubly robust (DR) estimator was derived and has shown its effectiveness in estimation, especially when the model assumption is violated. However, the current form of the DR estimator is restricted to binary treatments, while the treatment can be continuous in many real-world applications. The primary obstacle to continuous treatments resides in the delta function present in the original DR estimator, making it infeasible in causal effect estimation and introducing a heavy computational burden in nuisance function estimation. To address these challenges, we propose a kernel-based DR estimator that can well handle continuous treatments. Equipped with its smoothness, we show that its oracle form is a consistent approximation of the influence function. Further, we propose a new approach to efficiently solve the nuisance functions. We then provide a comprehensive convergence analysis in terms of the mean square error. We demonstrate the utility of our estimator on synthetic datasets and real-world applications.

stat.ME↗