arXiv subjects

Fei Yang

Publications and source records attributed to Fei Yang.

At least 37 records · Page 2Linked to original sources

arXiv2026

KernelSkill: A Multi-Agent Framework for GPU Kernel Optimization

Improving GPU kernel efficiency is crucial for advancing AI systems. Recent work has explored leveraging large language models (LLMs) for GPU kernel generation and optimization. However, existing LLM-based kernel optimization pipelines typically rely on opaque, implicitly learned heuristics within the LLMs to determine optimization strategies. This leads to inefficient trial-and-error and weakly interpretable optimizations. Our key insight is to replace implicit heuristics with expert optimization skills that are knowledge-driven and aware of task trajectories. Specifically, we present KernelSkill, a multi-agent framework with a dual-level memory architecture. KernelSkill operates by coordinating agents with long-term memory of reusable expert skills and short-term memory to prevent repetitive backtracking. On KernelBench Levels 1-3, KernelSkill achieves a 100% success rate and average speedups of 5.44x, 2.82x, and 1.92x over Torch Eager on Levels 1, 2, and 3, respectively, outperforming prior baselines. Code is available at https://github.com/0satan0/KernelMem/.

cs.LG↗

arXiv2026

A composite electron-lattice order: electronic nematicity of 2DEG and polarization density waves at a near-ferroelectric interface

We consider a two-dimensional electron gas (2DEG) formed at a near-ferroelectric interface and strongly coupled to polar phonons. Through a self-consistent microscopic many-body calculation, we show that the coupled system stabilizes a composite electron-lattice ordered state in which the lattice polarization spontaneously forms a polarization density wave (PDW), accompanied by an electronic stripe order in the 2DEG. This intertwined order partially reconstructs the electronic spectrum and generates a twofold quasiparticle anisotropy, giving rise to electronic nematicity at the single-particle level. However, under strong external electric fields, the nematic response becomes dominated by the collective sliding dynamics of the composite order: the sliding motion overwhelms the quasiparticle anisotropy and produces a strongly enhanced nematic signal with higher-order angular harmonics. The theory offers a natural explanation for several anomalous transport and anisotropic responses recently observed at the KTaO$_3$ (111) interface. We also estimate the mean-field transition temperature of this emergent ordered state, obtaining good agreement with experiments, and analyze its evolution with several tuning parameters. The proposed composite order, along with the field-induced crossover from quasiparticle-driven to sliding-dominated nematicity, provides a distinct mechanism of nematicity arising from many-body effects and collective dynamics in critical electron-boson systems, with applicability beyond ferroelectric platforms.

cond-mat.str-el↗

arXiv2026

Robust optimal reconciliation for hierarchical time series forecasting with M-estimation

Aggregation constraints, arising from geographical or sectoral division, frequently emerge in a large set of time series. Coherent forecasts of these constrained series are anticipated to conform to their hierarchical structure organized by the aggregation rules. To enhance its resilience against potential irregular series, we explore the robust reconciliation process for hierarchical time series (HTS) forecasting. We incorporate M-estimation to obtain the reconciled forecasts by minimizing a robust loss function of transforming a group of base forecasts subject to the aggregation constraints. The related minimization procedure is developed and implemented through a modified Newton-Raphson algorithm via local quadratic approximation. Extensive numerical experiments are carried out to evaluate the performance of the proposed method, and the results suggest its feasibility in handling numerous abnormal cases (for instance, series with non-normal errors). The proposed robust reconciliation also demonstrates excellent efficiency when no outliers exist in HTS. Finally, we showcase the practical application of the proposed method in a real-data study on Australian domestic tourism.

stat.AP↗

arXiv2026

LongSpeech: A Scalable Benchmark for Transcription, Translation and Understanding in Long Speech

Recent advances in audio-language models have demonstrated remarkable success on short, segment-level speech tasks. However, real-world applications such as meeting transcription, spoken document understanding, and conversational analysis require robust models capable of processing and reasoning over long-form audio. In this work, we present LongSpeech, a large-scale and scalable benchmark specifically designed to evaluate and advance the capabilities of speech models on long-duration audio. LongSpeech comprises over 100,000 speech segments, each approximately 10 minutes long, with rich annotations for ASR, speech translation, summarization, language detection, speaker counting, content separation, and question answering. We introduce a reproducible pipeline for constructing long-form speech benchmarks from diverse sources, enabling future extensions. Our initial experiments with state-of-the-art models reveal significant performance gaps, with models often specializing in one task at the expense of others and struggling with higher-level reasoning. These findings underscore the challenging nature of our benchmark. Our benchmark will be made publicly available to the research community.

cs.SD↗

arXiv2025

Marco-ASR: A Principled and Metric-Driven Framework for Fine-Tuning Large-Scale ASR Models for Domain Adaptation

Automatic Speech Recognition (ASR) models have achieved remarkable accuracy in general settings, yet their performance often degrades in domain-specific applications due to data mismatch and linguistic variability. This challenge is amplified for modern Large Language Model (LLM)-based ASR systems, whose massive scale and complex training dynamics make effective fine-tuning non-trivial. To address this gap, this paper proposes a principled and metric-driven fine-tuning framework for adapting both traditional and LLM-based ASR models to specialized domains. The framework emphasizes learning rate optimization based on performance metrics, combined with domain-specific data transformation and augmentation. We empirically evaluate our framework on state-of-the-art models, including Whisper, Whisper-Turbo, and Qwen2-Audio, across multi-domain, multilingual, and multi-length datasets. Our results not only validate the proposed framework but also establish practical protocols for improving domain-specific ASR performance while preventing overfitting.

cs.SD↗

arXiv2025

Sharpness-aware Dynamic Anchor Selection for Generalized Category Discovery

Generalized category discovery (GCD) is an important and challenging task in open-world learning. Specifically, given some labeled data of known classes, GCD aims to cluster unlabeled data that contain both known and unknown classes. Current GCD methods based on parametric classification adopt the DINO-like pseudo-labeling strategy, where the sharpened probability output of one view is used as supervision information for the other view. However, large pre-trained models have a preference for some specific visual patterns, resulting in encoding spurious correlation for unlabeled data and generating noisy pseudo-labels. To address this issue, we propose a novel method, which contains two modules: Loss Sharpness Penalty (LSP) and Dynamic Anchor Selection (DAS). LSP enhances the robustness of model parameters to small perturbations by minimizing the worst-case loss sharpness of the model, which suppressing the encoding of trivial features, thereby reducing overfitting of noise samples and improving the quality of pseudo-labels. Meanwhile, DAS selects representative samples for the unknown classes based on KNN density and class probability during the model training and assigns hard pseudo-labels to them, which not only alleviates the confidence difference between known and unknown classes but also enables the model to quickly learn more accurate feature distribution for the unknown classes, thus further improving the clustering accuracy. Extensive experiments demonstrate that the proposed method can effectively mitigate the noise of pseudo-labels, and achieve state-of-the-art results on multiple GCD benchmarks.

cs.CV↗

arXiv2025

SRNN: Spatiotemporal Relational Neural Network for Intuitive Physics Understanding

Human prowess in intuitive physics remains unmatched by machines. To bridge this gap, we argue for a fundamental shift towards brain-inspired computational principles. This paper introduces the Spatiotemporal Relational Neural Network (SRNN), a model that establishes a unified neural representation for object attributes, relations, and timeline, with computations governed by a Hebbian ``Fire Together, Wire Together'' mechanism across dedicated \textit{What} and \textit{How} pathways. This unified representation is directly used to generate structured linguistic descriptions of the visual scene, bridging perception and language within a shared neural substrate. On the CLEVRER benchmark, SRNN achieves competitive performance, thereby confirming its capability to represent essential spatiotemporal relations from the visual stream. Cognitive ablation analysis further reveals a benchmark bias, outlining a path for a more holistic evaluation. Finally, the white-box nature of SRNN enables precise pinpointing of error root causes. Our work provides a proof-of-concept that confirms the viability of translating key principles of biological intelligence into engineered systems for intuitive physics understanding in constrained environments.

cs.AI↗

arXiv2025

Dephasing-induced Quantum Hall Criticality in the Quantum Anomalous Hall system

Conventional wisdom holds that static disorder is indispensable to the integer quantum Hall effect, underpinning both quantized plateaus and the plateau-plateau transition. We show that pure dephasing, without elastic disorder, is sufficient to generate the same $θ$ driven criticality. Starting from a Keldysh formulation, we derive an open system nonlinear $σ$ model (NL$σ$M) for class A with a topological $θ$ term but no Cooperon sector, and we demonstrate that nonperturbative instantons still govern a two parameter flow of $(σ_{xx},σ_{xy})$. Evaluating $θ$ in a dephasing quantum anomalous Hall setting, we predict a quantum Hall critical point at $σ_{xy}=1/2$ with finite $σ_{xx}$ the hallmark of the integer quantum Hall universality class realized without Anderson localization. Boundary driven simulations of the Qi_Wu_Zhang model with local dephasing confirm this prediction and provide an experimentally aligned protocol to extract $(σ_{xx},σ_{xy})$ from Hall potential maps. By establishing dephasing as a self contained route to Hall criticality, our framework reframes plateau physics in open solid state and cold atom platforms and offers practical diagnostics for topological transport in nonunitary matter.

cond-mat.mes-hall↗

arXiv2025

Improving Video Generation with Human Feedback

Video generation has achieved significant advances through rectified flow techniques, but issues like unsmooth motion and misalignment between videos and prompts persist. In this work, we develop a systematic pipeline that harnesses human feedback to mitigate these problems and refine the video generation model. Specifically, we begin by constructing a large-scale human preference dataset focused on modern video generation models, incorporating pairwise annotations across multi-dimensions. We then introduce VideoReward, a multi-dimensional video reward model, and examine how annotations and various design choices impact its rewarding efficacy. From a unified reinforcement learning perspective aimed at maximizing reward with KL regularization, we introduce three alignment algorithms for flow-based models. These include two training-time strategies: direct preference optimization for flow (Flow-DPO) and reward weighted regression for flow (Flow-RWR), and an inference-time technique, Flow-NRG, which applies reward guidance directly to noisy videos. Experimental results indicate that VideoReward significantly outperforms existing reward models, and Flow-DPO demonstrates superior performance compared to both Flow-RWR and supervised fine-tuning methods. Additionally, Flow-NRG lets users assign custom weights to multiple objectives during inference, meeting personalized video quality needs.

cs.CV↗

arXiv2025

Logo-VGR: Visual Grounded Reasoning for Open-world Logo Recognition

Recent advances in multimodal large language models (MLLMs) have been primarily evaluated on general-purpose benchmarks, while their applications in domain-specific scenarios, such as intelligent product moderation, remain underexplored. To address this gap, we introduce an open-world logo recognition benchmark, a core challenge in product moderation. Unlike traditional logo recognition methods that rely on memorizing representations of tens of thousands of brands-an impractical approach in real-world settings-our proposed method, Logo-VGR, enables generalization to large-scale brand recognition with supervision from only a small subset of brands. Specifically, we reformulate logo recognition as a comparison-based task, requiring the model to match product images with candidate logos rather than directly generating brand labels. We further observe that existing models tend to overfit by memorizing brand distributions instead of learning robust multimodal reasoning, which results in poor performance on unseen brands. To overcome this limitation, Logo-VGR introduces a new paradigm of domain-specific multimodal reasoning: Logo Perception Grounding injects domain knowledge, and Logo-Guided Visual Grounded Reasoning enhances the model's reasoning capability. Experimental results show that Logo-VGR outperforms strong baselines by nearly 10 points in OOD settings, demonstrating superior generalization.

cs.CV↗

arXiv2025

Full-Dimensional Beamforming for Multi-User MIMO-OFDM ISAC for Low-Altitude UAV with Zero Sensing Resource Allocation

Low-altitude unmanned aerial vehicles (UAVs) are expected to play an important role for low-altitude economy with a wide range of applications like precise agriculture, aerial delivery and surveillance. Integrated sensing and communication (ISAC) is a key technology to enable the large-scale deployment and routine usage of UAVs by providing both communication and sensing services efficiently. For UAV ISAC systems, as UAV often acts as both a communication user equipment (UE) and a sensing target, traditional ISAC systems that usually allocate dedicated TF resources for sensing are inefficient due to the severe degradation of communication spectral efficiency. To address this issue, in this paper, we propose a novel multiple-input multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM)-based ISAC framework for UAVs that eliminates the need for dedicated sensing TF resources, achieving zero TF sensing overhead. By designing the transmit beamforming to meet the requirements for both communication and sensing tasks, our proposed approach enables the communication TF resources to be fully reused for sensing, thereby enhancing both the communication sum rate and the sensing performance in terms of resolution, unambiguous range, and accuracy. Additionally, we introduce a low-complexity target searching beamforming algorithm and a two-stage super-resolution sensing algorithm, which ensure efficient implementation. Simulation results demonstrate that the proposed MIMO-OFDM-ISAC framework not only improves the communication sum rate but also outperforms traditional ISAC systems in sensing performance, making it a promising solution for future ISAC systems to support low-altitude UAVs.

eess.SP↗

arXiv2025

Rational maps whose Julia sets are generalized Sierpiński gaskets

It has been shown that the Sierpiński gasket-like sets can appear as the Julia sets of some geometrically finite rational maps. In this paper we prove that such type of Julia sets can also appear in the rational maps containing Siegel disks, Cremer points or which are infinitely renormalizable. Based on this, we prove the existence of gasket Julia sets with positive area. Moreover, we present a criterion which guarantees the existence of gasket Julia sets in some rational maps having exactly one fixed attracting or parabolic basin.

math.DS↗

arXiv2025

VIVID-10M: A Dataset and Baseline for Versatile and Interactive Video Local Editing

Diffusion-based image editing models have made remarkable progress in recent years. However, achieving high-quality video editing remains a significant challenge. One major hurdle is the absence of open-source, large-scale video editing datasets based on real-world data, as constructing such datasets is both time-consuming and costly. Moreover, video data requires a significantly larger number of tokens for representation, which substantially increases the training costs for video editing models. Lastly, current video editing models offer limited interactivity, often making it difficult for users to express their editing requirements effectively in a single attempt. To address these challenges, this paper introduces a dataset VIVID-10M and a baseline model VIVID. VIVID-10M is the first large-scale hybrid image-video local editing dataset aimed at reducing data construction and model training costs, which comprises 9.7M samples that encompass a wide range of video editing tasks. VIVID is a Versatile and Interactive VIdeo local eDiting model trained on VIVID-10M, which supports entity addition, modification, and deletion. At its core, a keyframe-guided interactive video editing mechanism is proposed, enabling users to iteratively edit keyframes and propagate it to other frames, thereby reducing latency in achieving desired outcomes. Extensive experimental evaluations show that our approach achieves state-of-the-art performance in video local editing, surpassing baseline methods in both automated metrics and user studies. The VIVID-10M dataset are open-sourced at https://kwaivgi.github.io/VIVID/.

cs.CV↗

arXiv2025

Mind the Gap: Preserving and Compensating for the Modality Gap in CLIP-Based Continual Learning

Continual learning aims to enable models to learn sequentially from continuously incoming data while retaining performance on previously learned tasks. With the Contrastive Language-Image Pre-trained model (CLIP) exhibiting strong capabilities across various downstream tasks, there has been growing interest in leveraging CLIP for continual learning in such scenarios. Most existing works overlook the inherent modality gap in CLIP, a key factor in its generalization and adaptability. In this paper, we analyze the variations in the modality gap during the fine-tuning of vision-language pre-trained models. Our observations reveal that the modality gap effectively reflects the extent to which pre-trained knowledge is preserved. Based on these insights, we propose a simple yet effective method, MG-CLIP, that improves CLIP's performance in class-incremental learning. Our approach leverages modality gap preservation to mitigate forgetting and modality gap compensation to enhance the capacity for new data, introducing a novel modality-gap-based perspective for continual learning. Extensive experiments on multiple benchmarks demonstrate that our method outperforms existing approaches without requiring additional replay data. Our code is available at https://github.com/linlany/MindtheGap.

cs.CV↗

arXiv2025

Improving Continual Learning Performance and Efficiency with Auxiliary Classifiers

Continual learning is crucial for applying machine learning in challenging, dynamic, and often resource-constrained environments. However, catastrophic forgetting - overwriting previously learned knowledge when new information is acquired - remains a major challenge. In this work, we examine the intermediate representations in neural network layers during continual learning and find that such representations are less prone to forgetting, highlighting their potential to accelerate computation. Motivated by these findings, we propose to use auxiliary classifiers(ACs) to enhance performance and demonstrate that integrating ACs into various continual learning methods consistently improves accuracy across diverse evaluation settings, yielding an average 10% relative gain. We also leverage the ACs to reduce the average cost of the inference by 10-60% without compromising accuracy, enabling the model to return the predictions before computing all the layers. Our approach provides a scalable and efficient solution for continual learning.

cs.LG↗

arXiv2025

PropNet: a White-Box and Human-Like Network for Sentence Representation

Transformer-based embedding methods have dominated the field of sentence representation in recent years. Although they have achieved remarkable performance on NLP missions, such as semantic textual similarity (STS) tasks, their black-box nature and large-data-driven training style have raised concerns, including issues related to bias, trust, and safety. Many efforts have been made to improve the interpretability of embedding models, but these problems have not been fundamentally resolved. To achieve inherent interpretability, we propose a purely white-box and human-like sentence representation network, PropNet. Inspired by findings from cognitive science, PropNet constructs a hierarchical network based on the propositions contained in a sentence. While experiments indicate that PropNet has a significant gap compared to state-of-the-art (SOTA) embedding models in STS tasks, case studies reveal substantial room for improvement. Additionally, PropNet enables us to analyze and understand the human cognitive processes underlying STS benchmarks.

cs.CL↗

arXiv2025

Local connectivity of Julia sets of some transcendental entire functions with Siegel disks

Based on the weak expansion property of a long iteration of a family of quasi-Blaschke products near the unit circle established recently, we prove that the Julia sets of a number of transcendental entire functions with bounded type Siegel disks are locally connected. In particular, if $θ$ is of bounded type, then the Julia set of the sine function $S_θ(z)=e^{2πiθ}\sin(z)$ is locally connected. Moreover, we prove the existence of transcendental entire functions having Siegel disks and locally connected Julia sets with asymptotic values.

math.DS↗

arXiv2025

Not All Parameters Matter: Masking Diffusion Models for Enhancing Generation Ability

The diffusion models, in early stages focus on constructing basic image structures, while the refined details, including local features and textures, are generated in later stages. Thus the same network layers are forced to learn both structural and textural information simultaneously, significantly differing from the traditional deep learning architectures (e.g., ResNet or GANs) which captures or generates the image semantic information at different layers. This difference inspires us to explore the time-wise diffusion models. We initially investigate the key contributions of the U-Net parameters to the denoising process and identify that properly zeroing out certain parameters (including large parameters) contributes to denoising, substantially improving the generation quality on the fly. Capitalizing on this discovery, we propose a simple yet effective method-termed ``MaskUNet''- that enhances generation quality with negligible parameter numbers. Our method fully leverages timestep- and sample-dependent effective U-Net parameters. To optimize MaskUNet, we offer two fine-tuning strategies: a training-based approach and a training-free approach, including tailored networks and optimization functions. In zero-shot inference on the COCO dataset, MaskUNet achieves the best FID score and further demonstrates its effectiveness in downstream task evaluations. Project page: https://gudaochangsheng.github.io/MaskUnet-Page/

cs.CV↗