Searcharxiv⌕ Search

arXiv subjects

Feng Lu

Publications and source records attributed to Feng Lu.

At least 19 recordsLinked to original sources

Hy ASR 3.0 Technical Report

We present Hy ASR 3.0, a large language model (LLM)-based automatic speech recognition (ASR) system that delivers strong recognition accuracy, supports flexible contextual conditioning, and maintains robust recognition performance across diverse speech conditions. Hy ASR 3.0 supports Mandarin Chinese, English and 13 Chinese dialect groups, including Cantonese. Among the evaluated systems, it achieves the lowest aggregate error rates on public Mandarin Chinese, English, and Cantonese benchmarks and the lowest error rates on the corresponding internal test sets, alongside first-tier dialect recognition performance. Beyond audio-only transcription, Hy ASR 3.0 supports multiple forms of external context, including hotword lists, question--answer dialogues, and summaries, enabling the use of task-relevant information during recognition. The model also leverages intra-utterance context to disambiguate homophonic and phonetically similar keywords using information from other parts of the same utterance, without requiring a separate contextual input. Evaluations demonstrate strong contextual recognition in both external-context and intra-utterance settings, as well as robust recognition of long-form speech, children's and older adults' speech, low-volume speech, and speech in noisy environments. This report describes the model architecture and training methodology and presents evaluations across recognition accuracy, context utilization, and robustness.

cs.CL↗

Deviation-Aware Pretraining for Cardiovascular Disease Detection from Electrocardiograms

Electrocardiogram (ECG) signals are critical for cardiovascular disease (CVD) detection. Most current deep learning models rely on supervised, task-specific training, which limits their ability to generalize across diseases, patient populations, and acquisition settings. This paper presents a self-supervised ECG feature extraction framework that learns from nearly one million unlabeled 12-lead ECG recordings. The framework includes two modules: Deviation Localization Embedding (DLE), which detects waveform-level deviations from normal patterns, and Association Reconstruction Embedding (ARE), which models cross-lead associations through deviation-guided reconstruction to capture abnormal patterns expressed across multiple leads. These modules enable the framework to capture subtle deviation-related patterns and their cross-lead context. The framework achieves up to 6.5% AUROC and 12.8% F1 improvements over prior methods on multiple public datasets, even without additional fine-tuning the pretrained feature extractor. Embedding visualizations show that the learned features support meaningful disease grouping and anomaly identification. The approach reduces the need for labeled data and can support clinicians in diverse healthcare settings. The framework has potential to improve early detection and monitoring of cardiovascular diseases, contributing to more accessible and equitable heart care.

cs.LG↗

Where, What, Why, and Importance: Structured Defect Grounding for Text-to-Image Feedback

Despite generating increasingly photorealistic images, text-to-image (T2I) models still exhibit localized, subtle, and structurally complex failures. Diagnosing these failures requires instance-level feedback that answers where a defect occurs, what type it is, why it is defective, and its importance to overall image quality. While recent dense-feedback methods move beyond scalar supervision, their heatmap-centric representations still formulate diagnosis as pixel-field regression, making it difficult to localize variable-cardinality defects and bind semantic reasons to individual failures. To address this representation bottleneck, we propose Structured Defect Grounding (SDG), which casts T2I diagnosis as structured set prediction by modeling each defect as a (location, type, reason, importance) tuple. To make this formulation trainable and measurable, we introduce SDG-30K, a 30K-image dataset with box-grounded annotations across four modern T2I generators, together with a dedicated evaluation protocol, SDG-Eval. Building on this structured representation, we further present a diagnosis-to-alignment framework in which a Vision-Language Model (VLM) serves as the SDG detector, and BoxFlow-GRPO converts predicted defect sets into box-derived, importance-weighted spatial rewards for diffusion model alignment. Extensive experiments show that our SDG detector outperforms leading proprietary VLMs on structured defect grounding, while SDG-guided rewards consistently improve T2I alignment and support localized image refinement. These results establish SDG as a unified, instance-level interface for diagnosing, evaluating, and enhancing modern generative models.

cs.CV↗

HyBRA-EEG: Hyperbolic Global-Local Brain-Region Aggregation for Self-Supervised EEG Representation Learning

Electroencephalography (EEG) representation learning is challenged by noisy signals, inter-subject variability, and heterogeneous electrode montages in different datasets and recording settings. Recent self-supervised EEG pretraining methods improve transferability by converting heterogeneous recordings into unified token or sequence representations, but channel-aware designs may still encode montage- or dataset-specific regularities. We propose HyBRA, a self-supervised framework based on hyperbolic global-local brain-region aggregation. HyBRA maps heterogeneous channel observations into a fixed regional token space by combining electrode coordinates, region priors, and feature-dependent hyperbolic distances. A channel-evidence injection branch preserves fine-grained discriminative cues during aggregation. The resulting tokens are processed by a Region-Time Transformer and pretrained with masked reconstruction of latent features and region-level signals. HyBRA outperforms the compared methods in average balanced accuracy on six benchmark datasets. Under a controlled six-fold evaluation on OpenBMI-MI, HyBRA matches CSBrain in balanced accuracy, improves AUROC by 1.84 percentage points and AUPRC by 5.23 percentage points, and uses 96.8\% fewer parameters. Matched comparisons on TUAB and TUEV show higher balanced accuracy for hyperbolic matching than for Euclidean distance or cosine similarity.

cs.LG↗

SURE-EVAL: A Systematic and Unified Agentic Framework for Reproducible Evaluation

Audio and speech models are released rapidly, but reported scores often conflate model capability with deployment and evaluation choices. The same checkpoint can produce different predictions under different runtimes, hardware, decoding settings, or fallback policies. Even fixed predictions can receive different scores under different normalization and metric implementations. Existing speech benchmarks standardize selected datasets or scoring procedures, but rarely connect heterogeneous model onboarding, controlled inference, and versioned scoring in one executable workflow. We introduce SURE-EVAL, a Systematic and Unified Agentic framework for Reproducible Evaluation of audio and speech systems. A Tool Agent Workflow converts model releases into isolated, verified callable tools. A Main Agent Workflow commits task-specific inference and scoring protocols, then delegates all score-bearing operations to versioned deterministic programs. Each result retains its runtime, protocol, pipeline nodes, predictions, and audit artifacts. Across 18 public releases covering automatic speech recognition, text-to-speech, voice conversion, speaker diarization, speaker-attributed recognition, and multi-task audio understanding, a Codex-only baseline completes 12 models in one shot, while the same agent with the SURE-EVAL Tool Agent Workflow completes all 18. We also conduct unified evaluations over seven ASR test conditions and two TTS subsets. A protocol analysis of three TTS systems finds absolute differences of 0.02-0.52 points between paper-reported and unified results, with the direction varying by model and language. These results show that reproducible evaluation requires controlling both model execution and output scoring.

eess.AS↗

Learning from synthetic photorealistic raindrop for single image raindrop removal

Raindrops adhered to camera lens or windshield are inevitable in rainy scenes and can become an issue for many computer vision systems such as autonomous driving. Because raindrop appearance is affected by too many parameters, therefore it is unlikely to find an effective model based solution. Learning based methods are also problematic, because traditional learning method cannot properly model the complex appearance. Whereas deep learning method lacks sufficiently large and realistic training data. To solve it, in our work, we propose the first photo-realistic dataset of synthetic adherent raindrops for training. The rendering is physics based with consideration of the water dynamic, geometric and photometry. The dataset contains various types of rainy scenes and particularly the rainy driving scenes. Based on the modeling of raindrop imagery, we introduce a detection network which has the awareness of the raindrop refraction as well as its blurring. Based on that, we propose the removal network that can well recover the image structure. Rigorous experiments demonstrate the state-of-the-art performance of our proposed framework.

cs.CV↗

Shallow Queries, Mature Values: Depth-Asynchronous Self-Speculation for Looped Transformers

Looped Transformers reuse a shared block across recurrent depths, making autoregressive decoding expensive because every generated token requires many sequential recurrent passes. Self-speculative decoders reduce this cost by drafting at an early depth and verifying at full depth, but typically bind draft computation to prefix representations from the same recurrent depth. We find that queries and keys approach their final-depth representations earlier than values, and controlled prefix-channel interventions show that mature values substantially improve shallow draft predictions. Motivated by this asymmetry, we introduce Depth-Asynchronous Self-Speculation (DAS), which decouples the depth of draft computation from the depth of verified-prefix representations it reads. Its Mature-V primitive lets shallow queries retrieve full-depth prefix values without additional recurrent computation. We further develop DAS-Wave, which combines depth-asynchronous prefix reads with carried parallel refinement, progressive block growth, and an independent full-depth verifier. Across four recurrent-model checkpoints and mathematics and code workloads, DAS-Wave achieves 4.00--6.96$\times$ mean throughput speedup over paired full-depth autoregressive decoding in the same inference stack. These results identify prefix-information depth as an effective design axis for recurrent self-speculation.

cs.LG↗

Self-constrained Magnetic Reconnection of a Solar Flare Co-spatial with a Non-ejected Filament

For flares associated with eruptive filaments, the filament normally rises above the post-flare loops. Here, we report an unusual event in which the post-flare loops instead overlie a filament that remains non-ejected throughout the flare. This event was simultaneously captured by the Chinese H$α$ Solar Explorer (CHASE), the Solar Extreme-UltraViolet Imager (SUVI) onboard Fengyun-4C, the Hard X-ray Imager (HXI) onboard the Advanced Space-based Solar Observatory (ASO-S) and the Solar Dynamics Observatory (SDO). Extreme Ultraviolet (EUV) observations provided by FY4C/SUVI and SDO show that two hot loop systems form at the flare onset. The western end of the northern loop system undergoes a southwestward drifting motion with gradually increasing shear, whereas the southern loop system progressively converges toward and eventually overlies the northern one. Meanwhile, three-dimensional magnetic field extrapolations reveal an increase in both the twist of the filament-supporting magnetic structure and its poloidal magnetic field after the flare. Taken together, these observations suggest that the flare is driven by a self-constraining zipper-type magnetic reconnection, in which highly sheared arcades reconnect to form a series of overlying flare arcades and an underlying flux rope. This reconnection continuously strengthens the magnetic confinement of the filament, preventing its outward eruption, while facilitating substantial plasma drainage along the reconnected magnetic field revealed by the CHASE H$α$ spectroscopic observations.

astro-ph.SR↗

A Chip-scale Space-time Multiplexed Gaussian Boson Sampling Processor Beyond 10,000 Photons

Gaussian boson sampling (GBS) has emerged as a leading photonic paradigmfor demonstrating quantum computational advantage. Nevertheless, state-ofthe-art GBS setups face practical barriers including stringent optical alignment, phase instability, and limited programmability, which impede scalable engineering deployment. The chip-scale space-time multiplexed architecturepromises to resolve these constraints, yet it strongly demands wafer-scale chipcapabilities to simultaneously satisfy stringent requirements on low loss, highprecision and high-speed modulation. Here we report the first chip-scale spacetime multiplexed GBS system, monolithically integrating high-speed electrooptic modulators, on-chip delay lines, and a time-space multiplexed interferometric network on a thin-film lithium niobate chip, operating at a 4-GHz clockrate with detection events of up to 11,059 photons within 1 millisecond. Beyond benchmarking quantum advantage, we further reconfigure the photonichardware into a GBS-powered world model for modelling physical dynamics,which achieves lower prediction error with fewer trainable readout parameters compared with a classical echo state network (ESN) baseline. Our resultsvalidate the feasibility of our endeavor towards scalable photonic quantumhardware, and pave the way for the versatile programmable applications offuture GBS quantum systems.

quant-ph↗

Are All Tokens Necessary for Visual Place Recognition? An Empirical Study of Token Reduction for Efficient Inference

Recent visual place recognition (VPR) methods based on vision transformers, particularly foundation models, have achieved remarkable recognition performance. However, these models process all visual tokens throughout the entire network, resulting in substantial computational overhead, which hinders their deployment in real-time and resource-constrained scenarios. A natural question thus arises: are all visual tokens necessary for VPR? To answer this question, we present the first systematic benchmark of token reduction for efficient visual place recognition. Our benchmark comprehensively evaluates representative token pruning, token merging, and hybrid pruning-merging methods across multiple state-of-the-art VPR models and diverse benchmark datasets covering urban, suburban, and natural environments. We further investigate token reduction from multiple perspectives, including recognition performance under different reduction configurations, computational complexity, inference speed, qualitative visualization, and deployment efficiency on edge devices. Through extensive experiments and in-depth analysis, our benchmark reveals multiple important characteristics of token reduction in VPR and provides several practical insights into the trade-offs between accuracy and inference efficiency. For example, token reduction can reduce computational cost by up to 29\% and improve throughput by up to 44\%, while incurring less than 1\% degradation in recognition accuracy. Overall, this work establishes a comprehensive foundation for future research on token-efficient VPR and efficient visual retrieval systems. Our codes and models will be available at https://github.com/Tong-Jin01/TokenReduction4VPR

cs.CV↗

Selectivity Drives Efficiency: Dataset Pruning for Visual Place Recognition

Recent visual place recognition (VPR) studies have increasingly relied on large-scale datasets to train more robust and discriminative models. Although this trend significantly improves recognition performance, it also introduces substantial storage and training costs, especially when new architectures or training strategies need to be repeatedly developed and evaluated. Dataset pruning (DP) provides a promising way to improve data efficiency by retaining only informative training data. However, conventional DP methods mainly follow the sample-wise classification paradigm, which overlooks the relation-dependent training nature of VPR, where supervision is typically formed by image pairs rather than independent images. To address this issue, we propose a place-wise dataset pruning framework tailored for VPR. Instead of pruning individual images, our method treats each place as the basic pruning unit and introduces two complementary novel metrics, i.e., intra-place diversity (IPD) and inter-place similarity (IPS), to evaluate the training value of each place. By jointly considering these two metrics, our method ranks all places and constructs a compact yet informative coreset, thereby allowing the pruned dataset to still support the training of robust and discriminative VPR models. Extensive experiments demonstrate that our method consistently outperforms state-of-the-art DP baselines under different pruning ratios while reducing selection and training costs. Moreover, by pruning a merged dataset roughly 3.5$\times$ the size of GSV-Cities to a comparable scale, our coreset maintains highly competitive performance, achieving 94.5\% R@1 on MSLS-val and 97.0\% R@1 on Nordland with only NetVLAD. Codes will be made publicly available.

cs.CV↗

IFAR: Multi-Perspective and Multi-Level Causal Discovery with LLMs

Large language models (LLMs) have developed rapidly, and their reasoning capabilities have become a hot research topic. However, there is still limited exploration of abductive reasoning. The multi-perspective and multi-level of causes is one of the core challenges of abductive reasoning, which cannot be solved well by existing methods. We construct a specialized dataset named DeepAbduction, which is designed for tracing the causes of pollution and disease, addressing the lack of datasets in this field. We propose Inverse-Forward Abductive Reasoning (IFAR) framework for LLMs multi-perspective and multi-level abductive reasoning. IFAR is zero-shot and combines generalized backward reasoning with relation-by-relation forward verification. Experimental results show that IFAR achieves an improvement of approximately 40% in the F1 score compared to other methods under mainstream LLMs, while maintaining a balance between recall and precision. Furthermore, IFAR enhances the performance of non-reasoning LLMs to surpass LLMs which have been trained for reasoning, and remains effective when applied to the latter. Code will be released after the acceptance of our work.

cs.AI↗

The Quality-Utility Paradox: Why High-Reward Data Impairs Small Model Mathematical Reasoning

Knowledge distillation from powerful reasoning models is widely used to improve Small Language Models (SLMs) on mathematical reasoning, often assuming that traces with higher reward model scores provide more useful supervision. We identify a counterintuitive \textbf{Quality-Utility Paradox} in mathematical reasoning distillation. Data refined or synthesized by a stronger Oracle obtains higher perceived quality according to reward models, yet consistently underperforms traces generated by the SLM itself and selected through rejection sampling across Qwen2.5, LLaMA-3, and DeepSeek families. Our analysis shows that Oracle refinement couples logical repair with distributional drift away from the SLM's native reasoning distribution. This drift increases the learner's adaptation cost and can outweigh the benefit of improved reasoning logic. To test this mechanism, we introduce \textbf{Style-Aligned Refinement}, which preserves the native trajectory of the SLM while retaining logical repair from the Oracle. This intervention lowers adaptation cost and restores downstream utility. These findings suggest that effective mathematical reasoning distillation should jointly optimize perceived solution quality and learner-data compatibility, rather than relying solely on reward-model scores. The datasets and code are available at https://github.com/Dracoqhl/Quality-Utility-Paradox.

cs.AI↗

GazeOnce360: Fisheye-Based 360° Multi-Person Gaze Estimation with Global-Local Feature Fusion

We present GazeOnce360, a novel end-to-end model for multi-person gaze estimation from a single tabletop-mounted upward-facing fisheye camera. Unlike conventional approaches that rely on forward-facing cameras in constrained viewpoints, we address the underexplored setting of estimating the 3D gaze direction of multiple people distributed across a 360° scene from an upward fisheye perspective. To support research in this setting, we introduce MPSGaze360, a large-scale synthetic dataset rendered using Unreal Engine, featuring diverse multi-person configurations with accurate 3D gaze and eye landmark annotations. Our model tackles the severe distortion and perspective variation inherent in fisheye imagery by incorporating rotational convolutions and eye landmark supervision. To better capture fine-grained eye features crucial for gaze estimation, we propose a dual-resolution architecture that fuses global low-resolution context with high-resolution local eye regions. Experimental results demonstrate the effectiveness of each component in our model. This work highlights the feasibility and potential of fisheye-based 360° gaze estimation in practical multi-person scenarios. Project page: https://caizhuojiang.github.io/GazeOnce360/.

cs.CV↗

SIAgent: Spatial Interaction Agent via LLM-powered Eye-Hand Motion Intent Understanding in VR

Eye-hand coordinated interaction is becoming a mainstream interaction modality in Virtual Reality (VR) user interfaces.Current paradigms for this multimodal interaction require users to learn predefined gestures and memorize multiple gesture-task associations, which can be summarized as an ``Operation-to-Intent" paradigm. This paradigm increases users' learning costs and has low interaction error tolerance. In this paper, we propose SIAgent, a novel "Intent-to-Operation" framework allowing users to express interaction intents through natural eye-hand motions based on common sense and habits. Our system features two main components: (1) intent recognition that translates spatial interaction data into natural language and infers user intent, and (2) agent-based execution that generates an agent to execute corresponding tasks. This eliminates the need for gesture memorization and accommodates individual motion preferences with high error tolerance. We conduct two user studies across over 60 interaction tasks, comparing our method with two "Operation-to-Intent" techniques. Results show our method achieves higher intent recognition accuracy than gaze + pinch interaction (97.2% vs 93.1%) while reducing arm fatigue and improving usability, and user preference. Another study verifies the function of eye gaze and hand motion channels in intent recognition. Our work offers valuable insights into enhancing VR interaction intelligence through intent-driven design. Our source code and LLM prompts will be made available upon publication.

cs.HC↗

Requesting Expert Reasoning: Augmenting LLM Agents with Learned Collaborative Intervention

Large Language Model (LLM) based agents excel at general reasoning but often fail in specialized domains where success hinges on long-tail knowledge absent from their training data. While human experts can provide this missing knowledge, their guidance is often unstructured and unreliable, making its direct integration into an agent's plan problematic. To address this, we introduce AHCE (Active Human-Augmented Challenge Engagement), a framework for on-demand Human-AI collaboration. At its core, the Human Feedback Module (HFM) employs a learned policy to treat the human expert as an interactive reasoning tool. Extensive experiments in Minecraft demonstrate the framework's effectiveness, increasing task success rates by 32% on normal difficulty tasks and nearly 70% on highly difficult tasks, all with minimal human intervention. Our work demonstrates that successfully augmenting agents requires learning how to request expert reasoning, moving beyond simple requests for help.

cs.AI↗

Quantum Mpemba effect in quasiperiodic systems

We study a one-dimensional quasiperiodic tight-binding model with simultaneous off-diagonal (hopping) and diagonal (onsite) modulations. Using the inverse participation ratio and the wave-packet centroid, we construct localization-delocalization phase diagrams for both equilibrium and nonequilibrium steady states. We analyze the robustness of initial-state properties under dissipation and characterize dissipation-induced localization-delocalization transitions (and their reversals) in detail. Trace-distance dynamics provide evidence for a quantum Mpemba effect: states prepared farther from the steady state can relax faster than states initialized closer to it. We propose a starting-line hypothesis that explains the presence or absence of this effect across parameter regimes. These results advance the understanding of steady-state phase transitions and relaxation dynamics in dissipatively driven quasiperiodic systems.

cond-mat.dis-nn↗

Towards Implicit Aggregation: Robust Image Representation for Place Recognition in the Transformer Era

Visual place recognition (VPR) is typically regarded as a specific image retrieval task, whose core lies in representing images as global descriptors. Over the past decade, dominant VPR methods (e.g., NetVLAD) have followed a paradigm that first extracts the patch features/tokens of the input image using a backbone, and then aggregates these patch features into a global descriptor via an aggregator. This backbone-plus-aggregator paradigm has achieved overwhelming dominance in the CNN era and remains widely used in transformer-based models. In this paper, however, we argue that a dedicated aggregator is not necessary in the transformer era, that is, we can obtain robust global descriptors only with the backbone. Specifically, we introduce some learnable aggregation tokens, which are prepended to the patch tokens before a particular transformer block. All these tokens will be jointly processed and interact globally via the intrinsic self-attention mechanism, implicitly aggregating useful information within the patch tokens to the aggregation tokens. Finally, we only take these aggregation tokens from the last output tokens and concatenate them as the global representation. Although implicit aggregation can provide robust global descriptors in an extremely simple manner, where and how to insert additional tokens, as well as the initialization of tokens, remains an open issue worthy of further exploration. To this end, we also propose the optimal token insertion strategy and token initialization method derived from empirical studies. Experimental results show that our method outperforms state-of-the-art methods on several VPR datasets with higher efficiency and ranks 1st on the MSLS challenge leaderboard. The code is available at https://github.com/lu-feng/image.

cs.CV↗