arXiv subjects

Peiyao Wang

Publications and source records attributed to Peiyao Wang.

At least 19 recordsLinked to original sources

arXiv2026

Qwen-Audio-3.0-ASR Technical Report

In recent years, automatic speech recognition (ASR) has witnessed transformative advancements driven by three complementary paradigms: data scaling, model scaling, and deep integration with large language models (LLMs). However, bridging the gap between academic benchmark performance and real-world production utility remains a persistent challenge, particularly in handling diverse regional dialects, dynamic entities and hotwords, long-range contextual information, and disfluent spontaneous speech. In this report, we present Qwen-Audio-3.0-ASR, a Mixture-of-Experts (MoE) LLM-based ASR system designed to address these production demands through a unified, instruction-following framework. The model is built upon the Qwen backbone, and is trained on tens of millions of hours of large-scale speech data. Qwen-Audio-3.0-ASR supports transcription across 30 languages and 16 Chinese dialectal varieties spanning eight major dialect regions. Beyond multilingual and dialectal recognition, the model provides production-oriented capabilities including industry-domain entity recognition, hierarchical hotword customization, native single-pass transcription polishing, and long-audio contextual modeling. We further develop a dedicated streaming variant, Qwen-Audio-3.0-ASR-Streaming, for latency-sensitive applications. Extensive evaluations on Chinese, English, multilingual, and real-world industrial test sets demonstrate state-of-the-art or highly competitive recognition performance across a broad range of evaluation conditions, with strong performance relative to leading commercial and proprietary systems including GPT-4o Transcribe and Gemini 3.1 Pro.

cs.CL↗

arXiv2026

Dual-Layer Agentic Memory with Fast Write Routing and Slow Consolidation

Large language model (LLM) agents operate in dynamic environments where knowledge continuously evolves. Existing memory systems typically treat external memory as a monotonically growing repository, inevitably leading to retrieval degradation and increasing computational costs over time. We argue that the core challenge is not retrieval alone, but managing the knowledge lifecycle: deciding what to externalize, update, or ultimately internalize. Inspired by Complementary Learning Systems (CLS) theory in neuroscience, we propose Dual-Layer Agentic Memory, a framework that shifts memory management to the write phase through cost-aware epistemic routing and periodic parametric consolidation. Incoming information is categorized as non-write, write-new, or write-update, and routed through a small-to-large model cascade that minimizes routing overhead while filtering redundant memories. A subsequent write-back phase selectively consolidates high-value external memories into model parameters via supervised fine-tuning. Experiments demonstrate the dual efficiency of our approach: a 1.7B/8B cascade prunes up to 68% of redundant external memory while escalating fewer than 50% of inputs, yet retains over 98% of the downstream QA Exact Match (EM) achieved by an exhaustive retention baseline. We further show that periodic consolidation successfully internalizes external knowledge, allowing the router to adaptively suppress redundant writes as the model's epistemic boundaries evolve. Overall, our framework presents a unified paradigm for agent memory: selective externalization followed by selective internalization. Code and dataset will be released upon acceptance.

cs.CL↗

arXiv2026

Rethinking Generalization in Graph Neural Networks: A Structural Complexity Perspective

Graph neural networks (GNNs) have emerged as a fundamental tool for learning from graph-structured data, achieving strong performance across a wide range of applications. However, understanding their generalization capabilities remains challenging due to the complex structural dependencies inherent in such data. Existing generalization analyses largely follow the classical machine learning paradigm, focusing primarily on model complexity while overlooking the fundamental role of graph structure. Therefore, in this work, we systematically investigate this role by asking: does the graph structure actually influence generalization, and if so, by how much? To answer the first question and validate our intuition, we theoretically prove that incorporating more edges into the prediction process transforms the input representations to be overly accommodating to the output model, thereby inducing overfitting. To address the second question, we formulate a structural complexity measure based on the number of effective edges and derive a Rademacher complexity-based generalization bound. In doing so, we demonstrate that GNN generalization depends explicitly on structural complexity, alongside traditional parameter-dependent factors. Motivated by these theoretical findings, we propose a structural entropy regularization method. This approach controls structural complexity by regulating effective edges to balance underfitting and overfitting, ultimately improving the generalization performance of GNNs.

cs.LG↗

arXiv2026

Reinforcing Structured Chain-of-Thought for Video Understanding

Multi-modal Large Language Models (MLLMs) show promise in video understanding. However, their reasoning often suffers from thinking drift and weak temporal comprehension, even when enhanced by Reinforcement Learning (RL) techniques like Group Relative Policy Optimization (GRPO). Moreover, existing RL methods usually depend on Supervised Fine-Tuning (SFT), which requires costly Chain-of-Thought (CoT) annotation and multi-stage training, and enforces fixed reasoning paths, limiting MLLMs' ability to generalize and potentially inducing bias. To overcome these limitations, we introduce Summary-Driven Reinforcement Learning (SDRL), a novel single-stage RL framework that obviates the need for SFT by utilizing a Structured CoT format: Summarize -> Think -> Answer. SDRL introduces two self-supervised mechanisms integrated into the GRPO objective: 1) Consistency of Vision Knowledge (CVK) enforces factual grounding by reducing KL divergence among generated summaries; and 2) Dynamic Variety of Reasoning (DVR) promotes exploration by dynamically modulating thinking diversity based on group accuracy. This novel integration effectively balances alignment and exploration, supervising both the final answer and the reasoning process. Our method achieves state-of-the-art performance on seven public VideoQA datasets.

cs.CV↗

arXiv2025

Fun-ASR Technical Report

In recent years, automatic speech recognition (ASR) has witnessed transformative advancements driven by three complementary paradigms: data scaling, model size scaling, and deep integration with large language models (LLMs). However, LLMs are prone to hallucination, which can significantly degrade user experience in real-world ASR applications. In this paper, we present Fun-ASR, a large-scale, LLM-based ASR system that synergistically combines massive data, large model capacity, LLM integration, and reinforcement learning to achieve state-of-the-art performance across diverse and complex speech recognition scenarios. Moreover, Fun-ASR is specifically optimized for practical deployment, with enhancements in streaming capability, noise robustness, code-switching, hotword customization, and satisfying other real-world application requirements. Experimental results show that while most LLM-based ASR systems achieve strong performance on open-source benchmarks, they often underperform on real industry evaluation sets. Thanks to production-oriented optimizations, Fun-ASR achieves state-of-the-art performance on real application datasets, demonstrating its effectiveness and robustness in practical settings. The code and models are accessible at https://github.com/FunAudioLLM/Fun-ASR .

cs.CL↗

arXiv2025

PhysCorr: Dual-Reward DPO for Physics-Constrained Text-to-Video Generation with Automated Preference Selection

Recent advances in text-to-video generation have achieved impressive perceptual quality, yet generated content often violates fundamental principles of physical plausibility - manifesting as implausible object dynamics, incoherent interactions, and unrealistic motion patterns. Such failures hinder the deployment of video generation models in embodied AI, robotics, and simulation-intensive domains. To bridge this gap, we propose PhysCorr, a unified framework for modeling, evaluating, and optimizing physical consistency in video generation. Specifically, we introduce PhysicsRM, the first dual-dimensional reward model that quantifies both intra-object stability and inter-object interactions. On this foundation, we develop PhyDPO, a novel direct preference optimization pipeline that leverages contrastive feedback and physics-aware reweighting to guide generation toward physically coherent outputs. Our approach is model-agnostic and scalable, enabling seamless integration into a wide range of video diffusion and transformer-based backbones. Extensive experiments across multiple benchmarks demonstrate that PhysCorr achieves significant improvements in physical realism while preserving visual fidelity and semantic alignment. This work takes a critical step toward physically grounded and trustworthy video generation.

cs.CV↗

arXiv2025

XDIP: A Curated X-ray Absorption Spectrum Dataset for Iron-Containing Proteins

Earth-abundant iron is an essential metal in regulating the structure and function of proteins. This study presents the development of a comprehensive X-ray Absorption Spectroscopy (XAS) database focused on iron-containing proteins, addressing a critical gap in available high-quality annotated spectral data for iron-containing proteins. The database integrates detailed XAS spectra with their corresponding local structural data of proteins and enables direct comparison between spectral features and structural motifs. Utilizing a combination of manual curation and semi-automated data extraction techniques, we developed a comprehensive dataset via extensive literature review, ensuring the quality and accuracy of data, which contains 437 protein structures and 1954 XAS spectrums. Our methods included careful documentation and validation processes to ensure accuracy and reproducibility. This dataset not only centralizes information on iron-containing proteins but also supports advanced data-driven discoveries, such as machine learning, to predict and analyze protein structure and functions. This work underscores the potential of integrating detailed spectroscopic data with structural biology to advance the field of biological chemistry and catalysis.

q-bio.BM↗

arXiv2025

Spectra-to-Structure and Structure-to-Spectra Inference Across the Periodic Table

X-ray Absorption Spectroscopy (XAS) is a powerful technique for probing local atomic environments, yet its interpretation remains limited by the need for expert-driven analysis, computationally expensive simulations, and element-specific heuristics. Recent advances in machine learning have shown promise for accelerating XAS interpretation, but many existing models are narrowly focused on specific elements, edge types, or spectral regimes. In this work, we present XAStruct, a learning-based system capable of both predicting XAS spectra from crystal structures and inferring local structural descriptors from XAS input. XAStruct is trained on a large-scale dataset spanning over 70 elements across the periodic table, enabling generalization to a wide variety of chemistries and bonding environments. The framework includes the first machine learning approach for predicting neighbor atom types directly from XAS spectra, as well as a generalizable regression model for mean nearest-neighbor distance that requires no element-specific tuning. By combining deep neural networks for complex structure property mappings with efficient baseline models for simpler tasks, XAStruct offers a scalable and extensible solution for data-driven XAS analysis and local structure inference. The source code will be released upon paper acceptance.

cs.LG↗

arXiv2025

SVQA-R1: Reinforcing Spatial Reasoning in MLLMs via View-Consistent Reward Optimization

Spatial reasoning remains a critical yet underdeveloped capability in existing vision-language models (VLMs), especially for Spatial Visual Question Answering (Spatial VQA) tasks that require understanding relative positions, distances, and object configurations. Inspired by the R1 paradigm introduced in DeepSeek-R1, which enhances reasoning in language models through rule-based reinforcement learning (RL), we propose SVQA-R1, the first framework to extend R1-style training to spatial VQA. In particular, we introduce Spatial-GRPO, a novel group-wise RL strategy that constructs view-consistent rewards by perturbing spatial relations between objects, e.g., mirror flipping, thereby encouraging the model to develop a consistent and grounded understanding of space. Our model, SVQA-R1, not only achieves dramatically improved accuracy on spatial VQA benchmarks but also exhibits interpretable reasoning paths even without using supervised fine-tuning (SFT) data. Extensive experiments and visualization demonstrate the effectiveness of SVQA-R1 across multiple spatial reasoning benchmarks.

cs.CV↗

arXiv2025

Beyond Dielectrics: Interfacial Water Polarization Governs Graphene-Based Electrochemical Interfaces

Water molecules are traditionally regarded as passive dielectric media in electrochemical systems. In this work, we challenge this conventional perspective using molecular dynamics simulations and theoretical analysis. We show that interfacial water is polarized differently from bulk water and effectively screens the electrostatic potential between ions and the surface. This goes beyond the classic electric double layer (EDL) model, which treated water as merely a passive dielectric. The observed overscreening occurs because a significant portion of water polarization directly responds to the graphene surface, in addition to screening the electrostatic interactions between ions and charged surfaces. Furthermore, we reveal that this surface-induced polarization of interfacial water governs the electric potential distribution and EDL capacitance, and can even invert the electrode surface potential polarity, overriding the contribution of ions. These molecular-level insights lead to a revised EDL model that more accurately describes the electric and chemical potential distributions in the interfacial EDL regions.

physics.chem-ph↗

arXiv2024

Efficient Temporal Action Segmentation via Boundary-aware Query Voting

Although the performance of Temporal Action Segmentation (TAS) has improved in recent years, achieving promising results often comes with a high computational cost due to dense inputs, complex model structures, and resource-intensive post-processing requirements. To improve the efficiency while keeping the performance, we present a novel perspective centered on per-segment classification. By harnessing the capabilities of Transformers, we tokenize each video segment as an instance token, endowed with intrinsic instance segmentation. To realize efficient action segmentation, we introduce BaFormer, a boundary-aware Transformer network. It employs instance queries for instance segmentation and a global query for class-agnostic boundary prediction, yielding continuous segment proposals. During inference, BaFormer employs a simple yet effective voting strategy to classify boundary-wise segments based on instance segmentation. Remarkably, as a single-stage approach, BaFormer significantly reduces the computational costs, utilizing only 6% of the running time compared to state-of-the-art method DiffAct, while producing better or comparable accuracy over several popular benchmarks. The code for this project is publicly available at https://github.com/peiyao-w/BaFormer.

cs.CV↗

arXiv2024

CTC Blank Triggered Dynamic Layer-Skipping for Efficient CTC-based Speech Recognition

Deploying end-to-end speech recognition models with limited computing resources remains challenging, despite their impressive performance. Given the gradual increase in model size and the wide range of model applications, selectively executing model components for different inputs to improve the inference efficiency is of great interest. In this paper, we propose a dynamic layer-skipping method that leverages the CTC blank output from intermediate layers to trigger the skipping of the last few encoder layers for frames with high blank probabilities. Furthermore, we factorize the CTC output distribution and perform knowledge distillation on intermediate layers to reduce computation and improve recognition accuracy. Experimental results show that by utilizing the CTC blank, the encoder layer depth can be adjusted dynamically, resulting in 29% acceleration of the CTC model inference with minor performance degradation.

eess.AS↗

arXiv2023

DIR-AS: Decoupling Individual Identification and Temporal Reasoning for Action Segmentation

Fully supervised action segmentation works on frame-wise action recognition with dense annotations and often suffers from the over-segmentation issue. Existing works have proposed a variety of solutions such as boundary-aware networks, multi-stage refinement, and temporal smoothness losses. However, most of them take advantage of frame-wise supervision, which cannot effectively tackle the evaluation metrics with different granularities. In this paper, for the desirable large receptive field, we first develop a novel local-global attention mechanism with temporal pyramid dilation and temporal pyramid pooling for efficient multi-scale attention. Then we decouple two inherent goals in action segmentation, ie, (1) individual identification solved by frame-wise supervision, and (2) temporal reasoning tackled by action set prediction. Afterward, an action alignment module fuses these different granularity predictions, leading to more accurate and smoother action segmentation. We achieve state-of-the-art accuracy, eg, 82.8% (+2.6%) on GTEA and 74.7% (+1.2%) on Breakfast, which demonstrates the effectiveness of our proposed method, accompanied by extensive ablation studies. The code will be made available later.

cs.CV↗

arXiv2022

Joint and Individual Component Regression

Multi-group data are commonly seen in practice. Such data structure consists of data from multiple groups and can be challenging to analyze due to data heterogeneity. We propose a novel Joint and Individual Component Regression (JICO) model to analyze multi-group data. In particular, our proposed model decomposes the response into shared and group-specific components, which are driven by low-rank approximations of joint and individual structures from the predictors respectively. The joint structure has the same regression coefficients across multiple groups, whereas individual structures have group-specific regression coefficients. Moreover, the choice of global and individual ranks allows our model to cover global and group-specific models as special cases. For model estimation, we formulate this framework under the representation of latent components and propose an iterative algorithm to solve for the joint and individual scores under the new representation. To construct the latent scores, we utilize the Continuum Regression (CR), which provides a unified framework that covers the Ordinary Least Squares (OLS), the Partial Least Squares (PLS), and the Principal Component Regression (PCR) as its special cases. We show that JICO attains a good balance between global and group-specific models and remains flexible due to the usage of CR. Finally, we conduct simulation studies and analysis of an Alzheimer's disease dataset to further demonstrate the effectiveness of JICO. R implementation of JICO is available online at https://github.com/peiyaow/JICO.

stat.ME↗

arXiv2022

A Simple and Efficient Lattice Summation Method for Metallic Electrodes in Constant Potential Molecular Dynamics Simulation

The constant potential molecular dynamics simulation method proposed by Siepmann and Sprik and reformulated later by Reed (SR-CPM) has been widely employed to investigate the metallic electrolyte/electrode interfaces, especially for conducting nanochannels with complex connectivity, *e.g.*, carbide-derived carbon or graphene-assembled membrane. This work makes substantial extensions of this seminal SR-CPM approach. First, we introduce two numerical techniques to determine electrode atom charges with an order of magnitude improvement in computational efficiency compared with those widely employed methods. The first numerical technique dramatically accelerates the to calculation of the Ewald interaction matrix $\mathbf{E}$, which takes advantage of the existing highly optimised electrostatic codes. The second technique introduces a new preconditioning technique in the conjugate gradient method to considerably increase the computational efficiency of a linear equation system that determines electrode atomic charges. Our improved SR-CPM implemented in the LAMMPS package can handle extra-large systems, *e.g.*, over 8.1 million electrode atoms. Second, after demonstrating the importance of the electroneutrality constraint, we propose a two-step method to enforce electroneutrality in the following post-treatment step, applicable for matrix and iterative techniques. Third, we propose a solid theoretical analysis for the adjustable parameter $α_i$ (namely the atomic Hubbard-U $U_i^0$), which is arbitrarily selected in many SR-CPM simulation practices. We proposed that the optimised $α_i$ or $U_i^0$ should compensate for the electrical potential/energy discrepancy between the discrete atomistic model and the continuum limit. The analytical and optimal $α_i^0$ values are derived for a series of 2D materials.

physics.chem-ph↗

arXiv2022

DialAug: Mixing up Dialogue Contexts in Contrastive Learning for Robust Conversational Modeling

Retrieval-based conversational systems learn to rank response candidates for a given dialogue context by computing the similarity between their vector representations. However, training on a single textual form of the multi-turn context limits the ability of a model to learn representations that generalize to natural perturbations seen during inference. In this paper we propose a framework that incorporates augmented versions of a dialogue context into the learning objective. We utilize contrastive learning as an auxiliary objective to learn robust dialogue context representations that are invariant to perturbations injected through the augmentation method. We experiment with four benchmark dialogue datasets and demonstrate that our framework combines well with existing augmentation methods and can significantly improve over baseline BERT-based ranking architectures. Furthermore, we propose a novel data augmentation method, ConMix, that adds token level perturbations through stochastic mixing of tokens from other contexts in the batch. We show that our proposed augmentation method outperforms previous data augmentation approaches, and provides dialogue representations that are more robust to common perturbations seen during inference.

cs.CL↗

arXiv2022

The RoyalFlush System of Speech Recognition for M2MeT Challenge

This paper describes our RoyalFlush system for the track of multi-speaker automatic speech recognition (ASR) in the M2MeT challenge. We adopted the serialized output training (SOT) based multi-speakers ASR system with large-scale simulation data. Firstly, we investigated a set of front-end methods, including multi-channel weighted predicted error (WPE), beamforming, speech separation, speech enhancement and so on, to process training, validation and test sets. But we only selected WPE and beamforming as our frontend methods according to their experimental results. Secondly, we made great efforts in the data augmentation for multi-speaker ASR, mainly including adding noise and reverberation, overlapped speech simulation, multi-channel speech simulation, speed perturbation, front-end processing, and so on, which brought us a great performance improvement. Finally, in order to make full use of the performance complementary of different model architecture, we trained the standard conformer based joint CTC/Attention (Conformer) and U2++ ASR model with a bidirectional attention decoder, a modification of Conformer, to fuse their results. Comparing with the official baseline system, our system got a 12.22% absolute Character Error Rate (CER) reduction on the validation set and 12.11% on the test set.

cs.SD↗

arXiv2021

Quadrupling the stored charge by extending the accessible density of states

Nanosized energy storage, energy-harvesting, and functional devices are the three key components for integrated self-power systems. Here, we report on nanoscale electrochemical devices with a nearly three-fold enhanced stored charge under the field effect. We demonstrated the field-effect transistor can not only work as a functional component in nanodevices but also serve as an amplifier for the nanosized energy storage blocks. This unusual increase in energy storage is attributed to having a wide range of accessible electronic density of states (EDOS), hence redox reactions are occurring within the nanowire and not being confined to the surface. Initial results with MoS2 suggest that this field effect modulated energy storage mechanism may also apply to many other redox-active materials. Our work demonstrates the novel application of the field-effect in energy storage devices as a universal strategy to improve ion diffusion and the surface-active ion concentration of the active material, which can greatly enhance the charge storage ability of nanoscale devices. The fabrication process of the field-effect energy storage device is also compatible with microtechnology and can be integrated into other microdevices and nanodevices.

physics.app-ph↗