arXiv subjects

Yifei Wang

Publications and source records attributed to Yifei Wang.

At least 91 records · Page 5Linked to original sources

arXiv2026

CSRv2: Unlocking Ultra-Sparse Embeddings

In the era of large foundation models, the quality of embeddings has become a central determinant of downstream task performance and overall system capability. Yet widely used dense embeddings are often extremely high-dimensional, incurring substantial costs in storage, memory, and inference latency. To address these, Contrastive Sparse Representation (CSR) is recently proposed as a promising direction, mapping dense embeddings into high-dimensional but k-sparse vectors, in contrast to compact dense embeddings such as Matryoshka Representation Learning (MRL). Despite its promise, CSR suffers severe degradation in the ultra-sparse regime, where over 80% of neurons remain inactive, leaving much of its efficiency potential unrealized. In this paper, we introduce CSRv2, a principled training approach designed to make ultra-sparse embeddings viable. CSRv2 stabilizes sparsity learning through progressive k-annealing, enhances representational quality via supervised contrastive objectives, and ensures end-to-end adaptability with full backbone finetuning. CSRv2 reduces dead neurons from 80% to 20% and delivers a 14% accuracy gain at k=2, bringing ultra-sparse embeddings on par with CSR at k=8 and MRL at 32 dimensions, all with only two active features. While maintaining comparable performance, CSRv2 delivers a 7x speedup over MRL, and yields up to 300x improvements in compute and memory efficiency relative to dense embeddings in text representation. Extensive experiments across text and vision demonstrate that CSRv2 makes ultra-sparse embeddings practical without compromising performance, where CSRv2 achieves 7%/4% improvement over CSR when k=4 and further increases this gap to 14%/6% when k=2 in text/vision representation. By making extreme sparsity viable, CSRv2 broadens the design space for real-time and edge-deployable AI systems where both embedding quality and efficiency are critical.

cs.LG↗

arXiv2026

Benchmarking LLM Summaries of Multimodal Clinical Time Series for Remote Monitoring

Large language models (LLMs) can generate fluent clinical summaries of remote therapeutic monitoring time series. However, it remains unclear whether these narratives faithfully capture clinically significant events, such as sustained abnormalities. Existing evaluation metrics primarily focus on semantic similarity and linguistic quality, leaving event-level correctness largely unmeasured. To address this gap, we introduce an event-based evaluation framework for multimodal time-series summarization using the Technology-Integrated Health Management (TIHM)-1.5 dementia monitoring dataset. Clinically grounded daily events are derived through rule-based abnormal thresholds and temporal persistence criteria. Model-generated summaries are then aligned with these structured facts. Our evaluation protocol measures abnormality recall, duration recall, measurement coverage, and hallucinated event mentions. We benchmark three approaches: zero-shot prompting, statistical prompting, and a vision-based pipeline that uses rendered time-series visualizations. The results reveal a striking decoupling between conventional metrics and clinical event fidelity. Models that achieve high semantic similarity scores often exhibit near-zero abnormality recall. In contrast, the vision-based approach demonstrates the strongest event alignment, achieving 45.7% abnormality recall and 100% duration recall. These findings underscore the importance of event-aware evaluation to ensure reliable clinical time-series summarization.

cs.AI↗

arXiv2026

FASA: Frequency-aware Sparse Attention

The deployment of Large Language Models (LLMs) faces a critical bottleneck when handling lengthy inputs: the prohibitive memory footprint of the Key Value (KV) cache. To address this bottleneck, the token pruning paradigm leverages attention sparsity to selectively retain a small, critical subset of tokens. However, existing approaches fall short, with static methods risking irreversible information loss and dynamic strategies employing heuristics that insufficiently capture the query-dependent nature of token importance. We propose FASA, a novel framework that achieves query-aware token eviction by dynamically predicting token importance. FASA stems from a novel insight into RoPE: the discovery of functional sparsity at the frequency-chunk (FC) level. Our key finding is that a small, identifiable subset of "dominant" FCs consistently exhibits high contextual agreement with the full attention head. This provides a robust and computationally free proxy for identifying salient tokens. Building on this insight, FASA first identifies a critical set of tokens using dominant FCs, and then performs focused attention computation solely on this pruned subset. Across a spectrum of long-context tasks, from sequence modeling to complex CoT reasoning, FASA consistently outperforms all token-eviction baselines and achieves near-oracle accuracy, demonstrating remarkable robustness even under constraint budgets. Notably, on LongBench-V1, FASA reaches nearly 100\% of full-KV performance when only keeping 256 tokens, and achieves 2.56$\times$ speedup using just 18.9\% of the cache on AIME24.

cs.CL↗

arXiv2026

Unveiling the Role of ChatGPT in Software Development: Insights from Developer-ChatGPT Interactions on GitHub

The advent of Large Language Models (LLMs) has introduced a new paradigm in Software Engineering (SE), with generative AI tools like ChatGPT gaining widespread adoption among developers. While ChatGPT's potential has been extensively discussed, empirical evidence about how developers actually use LLMs' assistance in real-world practices remains limited. To bridge this gap, we conducted a large-scale empirical analysis of ChatGPT usage on GitHub, and we presented DevChat, a curated dataset of 2,547 publicly shared ChatGPT conversation links collected from GitHub between May 2023 and June 2024. Through comprehensively analyzing DevChat, we explored the characteristics of developer-ChatGPT interaction patterns and identified five key categories of developers' purposes for sharing developer-ChatGPT conversations during software development. Additionally, we investigated the dominant development-related activities in which ChatGPT is used, and presented a mapping framework that links GitHub data sources, development-related activities, and SE tasks. The findings show that interactions are typically short and task-focused (most are 1-3 turns); developers share conversations mainly to delegate tasks, resolve problems, and acquire knowledge, revealing five purpose categories; ChatGPT is most frequently engaged for Software Implementation and Maintenance & Evolution; we identified 39 fine-grained SE tasks supported by ChatGPT, with Code Generation & Completion as well as Code modification & Optimization being the most prominent. Our study offers a comprehensive mapping of ChatGPT's applications in real-world software development scenarios and provides a foundation for understanding LLMs' practical roles in software development.

cs.SE↗

arXiv2026

Biologically Plausible Learning via Bidirectional Spike-Based Distillation

Developing biologically plausible learning algorithms that can achieve performance comparable to error backpropagation remains a longstanding challenge. Existing approaches often compromise biological plausibility by entirely avoiding the use of spikes for error propagation or relying on both positive and negative learning signals, while the question of how spikes can represent negative values remains unresolved. To address these limitations, we introduce Bidirectional Spike-based Distillation (BSD), a novel learning algorithm that jointly trains a feedforward and a backward spiking network. We formulate learning as a transformation between two spiking representations (i.e., stimulus encoding and concept encoding) so that the feedforward network implements perception and decision-making by mapping stimuli to actions, while the backward network supports memory recall by reconstructing stimuli from concept representations. Extensive experiments on diverse benchmarks, including image recognition, image generation, and sequential regression, show that BSD achieves performance comparable to networks trained with classical error backpropagation. These findings represent a significant step toward biologically grounded, spike-driven learning in neural networks. Our code is available at https://github.com/alden199/Bidirectional-Spike-Based-Distillation.

cs.NE↗

arXiv2026

Overcoming sensitivity-bandwidth trade-off in mid-infrared spectroscopy by a microresonator-anchored swept laser

Optical frequency combs have revolutionized high-precision spectroscopy, yet an intrinsic trade-off between spectroscopic signal-to-noise ratio (sSNR) and measurement bandwidth ($B$) fundamentally constrains sensitive, broadband measurements. While broadband swept lasers offer a potential solution, generating broadband, ultrafast and linearly sweeping lasers with a narrow linewidth remains a significant challenge, particularly in the fingerprint mid-infrared (mid-IR) band. Here we overcome this limitation by using a microresonator-anchored ultrafast sweeping Fourier domain mode-locked (FDML) laser for mid-IR spectroscopy. We introduce a dual-microresonator-anchor approach: a microcomb provides frequency calibration and a high-Q microresonator resolves the instantaneous FDML lasing lineshape. The strategy enables accurate correction of the FDML laser's sweep nonlinearity and broad linewidth in the near-IR, allowing the FDML laser to function as a high-fidelity mid-IR light via difference frequency generation. The system achieves a record sSNR$\times$$B$ of 1.3$\times$10$^5$ THz$\cdot \sqrt{\rm Hz}$ and methane sensing precision of 9 ppb$\cdot$m$\cdot$$\sqrt{\rm s}$, while retaining GHz resolution to distinguish methane isotope. We further demonstrate broadband, coherent swept laser phase spectroscopy in the mid-IR, tolerating losses up to 78 dB. This work leverages advances in integrated photonics to overcome the fundamental limitations of precision spectroscopy, paving the way for next-generation, broadband, and ultra-sensitive mid-IR spectroscopic sensing systems.

physics.optics↗

arXiv2026

Gap Labelling for Almost Periodic Sturm-Liouville Operators

In this paper, we introduce a rotation number for almost periodic Sturm-Liouville operators in the spirit of Johnson and Moser. We then prove the gap labelling theorem in terms of rotation numbers for the operator in question. To do this, we rigorously prove the almost periodicity of Green's functions.

math.SP↗

arXiv2026

Autoregressive Models Rival Diffusion Models at ANY-ORDER Generation

Diffusion language models enable any-order generation and bidirectional conditioning, offering appealing flexibility for tasks such as infilling, rewriting, and self-correction. However, their formulation-predicting one part of a sequence from another within a single-step dependency-limits modeling depth and often yields lower sample quality and stability than autoregressive (AR) models. To address this, we revisit autoregressive modeling as a foundation and reformulate diffusion-style training into a structured multi-group prediction process. We propose Any-order Any-subset Autoregressive modeling (A3), a generalized framework that extends the standard AR factorization to arbitrary token groups and generation orders. A3 preserves the probabilistic rigor and multi-layer dependency modeling of AR while inheriting diffusion models' flexibility for parallel and bidirectional generation. We implement A3 through a two-stream attention architecture and a progressive adaptation strategy that transitions pretrained AR models toward any-order prediction. Experiments on question answering, commonsense reasoning, and story infilling demonstrate that A3 outperforms diffusion-based models while maintaining flexible decoding. This work offers a unified approach for a flexible, efficient, and novel language modeling paradigm.

cs.CL↗

arXiv2026

Hybrid RIS-Aided Digital Over-the-Air Computing for Edge AI Inference: Joint Feature Quantization and Active-Passive Beamforming Design

The vision of 6G networks aims to enable edge inference by leveraging ubiquitously deployed artificial intelligence (AI) models, facilitating intelligent environmental perception for a wide range of applications. A critical operation in edge inference is for an edge node (EN) to aggregate multi-view sensory features extracted by distributed agents, thereby boosting perception accuracy. Over-the-air computing (AirComp) emerges as a promising technique for rapid feature aggregation by exploiting the waveform superposition property of analog-modulated signals, which is, however, incompatible with existing digital communication systems. Meanwhile, hybrid reconfigurable intelligent surface (RIS), a novel RIS architecture capable of simultaneous signal amplification and reflection, exhibits potential for enhancing AirComp. Therefore, this paper proposes a Hybrid RIS-aided Digital AirComp (HRD-AirComp) scheme, which employs vector quantization to map high-dimensional features into discrete codewords that are digitally modulated into symbols for wireless transmission. By judiciously adjusting the AirComp transceivers and hybrid RIS reflection to control signal superposition across agents, the EN can estimate the aggregated features from the received signals. To endow HRD-AirComp with a task-oriented design principle, we derive a surrogate function for inference accuracy that characterizes the impact of feature quantization and over-the-air aggregation. Based on this surrogate, we formulate an optimization problem targeting inference accuracy maximization, and develop an efficient algorithm to jointly optimize the quantization bit allocation, agent transmission coefficients, EN receiving beamforming, and hybrid RIS reflection beamforming. Experimental results demonstrate that the proposed HRD-AirComp outperforms baselines in terms of both inference accuracy and uncertainty.

cs.NI↗

arXiv2026

Nixie: Efficient, Transparent Temporal Multiplexing for Consumer GPUs

Consumer machines are increasingly running large ML workloads such as large language models (LLMs), text-to-image generation, and interactive image editing. Unlike datacenter GPUs, consumer GPUs serve single-user, rapidly changing workloads, and each model's working set often nearly fills the GPU memory. As a result, existing sharing mechanisms (e.g., NVIDIA Unified Virtual Memory) perform poorly due to memory thrashing and excessive use of CPU pinned memory when multiple applications are active. We design and implement Nixie, a system that enables efficient and transparent temporal multiplexing on consumer GPUs without requiring any application or driver changes. Nixie is a system service that coordinates GPU memory allocation and kernel launch behavior to efficiently utilize the CPU-GPU bi-directional bandwidth and CPU pinned memory. A lightweight scheduler in Nixie further improves responsiveness by automatically prioritizing latency-sensitive interactive jobs using MLFQ-inspired techniques. Our evaluations show that Nixie improves latency of real interactive code-completion tasks by up to $3.8\times$ and saves up to 66.8% CPU pinned memory usage given the same latency requirement.

cs.OS↗

arXiv2026

On Generalized Strong and Norm Resolvent Convergence

We present a streamlined approach for generalized strong and norm convergence of self-adjoint operators in different Hilbert spaces. In particular, we establish convergence of associated (semi-)groups, (essential) spectra and spectral projections. In addition, we give some applications to Sturm-Liouville operators.

math.SP↗

arXiv2026

High-Q AlN microresonators for nonlinear near-infrared and near-visible photonics

High Q-factors of microresonators are crucial for nonlinear integrated photonics, as many nonlinear dynamics have quadratic or even cubic dependence on Q-factors. The unique material properties make AlN microresonators invaluable for microcomb generation, Raman lasing and visible integrated photonics. However, the loss level of AlN falls behind other integrated platforms. By optimizing the fabrication, we demonstrate record Q-factors of 5.4$\times$10$^6$ and 2.2$\times$10$^6$ for AlN microresonators in the near-infrared and near-visible, respectively. Polarized-mode-interaction was used to create anomalous dispersion to support bright AlN Dirac solitons. Measurement of polarization-dependent spectra reveals the polarization hybridization of the Dirac soliton. In a microresonator with normal dispersion, Raman assisted four-wave-mixing (RFWM) was observed to initiate platicon formation, adding an approach to generate normal dispersion microcombs. A design of width-varying waveguides was used to ensure both efficient coupling and high Q-factor for racetrack microresonators at 780 nm. The microresonator was pumped to generate near-visble Raman laser at 820 nm with a fundamental linewidth narrower than 220 Hz. Our work unlocks new opportunities for integrated AlN photonics by improving Q-factors and uncovering nonlinear dynamics in AlN microresonators.

physics.optics↗

arXiv2025

StreamEQA: Towards Streaming Video Understanding for Embodied Scenarios

As embodied intelligence advances toward real-world deployment, the ability to continuously perceive and reason over streaming visual inputs becomes essential. In such settings, an agent must maintain situational awareness of its environment, comprehend the interactions with surrounding entities, and dynamically plan actions informed by past observations, current contexts, and anticipated future events. To facilitate progress in this direction, we introduce StreamEQA, the first benchmark designed for streaming video question answering in embodied scenarios. StreamEQA evaluates existing MLLMs along two orthogonal dimensions: Embodied and Streaming. Along the embodied dimension, we categorize the questions into three levels: perception, interaction, and planning, which progressively assess a model's ability to recognize fine-grained visual details, reason about agent-object interactions, and perform high-level goal-directed reasoning. For the streaming dimension, questions are divided into backward, real-time, and forward reasoning, with each mode relying on a distinct temporal context. Built upon 156 independent long videos, StreamEQA defines 42 tasks and generates approximately 21K question-answer pairs with precise timestamps through a hybrid pipeline combining automated generation and human refinement. Evaluations of 13 state-of-the-art video-LLMs reveal that, despite strong performance on conventional benchmarks, these models still struggle with streaming video understanding in embodied scenarios. We hope StreamEQA will catalyze research on streaming video understanding for embodied applications.

cs.CV↗

arXiv2025

Large Language Model based Smart Contract Auditing with LLMBugScanner

This paper presents LLMBugScanner, a large language model (LLM) based framework for smart contract vulnerability detection using fine-tuning and ensemble learning. Smart contract auditing presents several challenges for LLMs: different pretrained models exhibit varying reasoning abilities, and no single model performs consistently well across all vulnerability types or contract structures. These limitations persist even after fine-tuning individual LLMs. To address these challenges, LLMBugScanner combines domain knowledge adaptation with ensemble reasoning to improve robustness and generalization. Through domain knowledge adaptation, we fine-tune LLMs on complementary datasets to capture both general code semantics and instruction-guided vulnerability reasoning, using parameter-efficient tuning to reduce computational cost. Through ensemble reasoning, we leverage the complementary strengths of multiple LLMs and apply a consensus-based conflict resolution strategy to produce more reliable vulnerability assessments. We conduct extensive experiments across multiple popular LLMs and compare LLMBugScanner with both pretrained and fine-tuned individual models. Results show that LLMBugScanner achieves consistent accuracy improvements and stronger generalization, demonstrating that it provides a principled, cost-effective, and extensible framework for smart contract auditing.

cs.CR↗

arXiv2025

Chart2Code-MoLA: Efficient Multi-Modal Code Generation via Adaptive Expert Routing

Chart-to-code generation is a critical task in automated data visualization, translating complex chart structures into executable programs. While recent Multi-modal Large Language Models (MLLMs) improve chart representation, existing approaches still struggle to achieve cross-type generalization, memory efficiency, and modular design. To address these challenges, this paper proposes C2C-MoLA, a multimodal framework that synergizes Mixture of Experts (MoE) with Low-Rank Adaptation (LoRA). The MoE component uses a complexity-aware routing mechanism with domain-specialized experts and load-balanced sparse gating, dynamically allocating inputs based on learnable structural metrics like element count and chart complexity. LoRA enables parameter-efficient updates for resource-conscious tuning, further supported by a tailored training strategy that aligns routing stability with semantic accuracy. Experiments on Chart2Code-160k show that the proposed model improves generation accuracy by up to 17%, reduces peak GPU memory by 18%, and accelerates convergence by 20%, when compared to standard fine-tuning and LoRA-only baselines, particularly on complex charts. Ablation studies validate optimal designs, such as 8 experts and rank-8 LoRA, and confirm scalability for real-world multimodal code generation.

cs.SE↗

arXiv2025

Monet: Reasoning in Latent Visual Space Beyond Images and Language

"Thinking with images" has emerged as an effective paradigm for advancing visual reasoning, extending beyond text-only chains of thought by injecting visual evidence into intermediate reasoning steps. However, existing methods fall short of human-like abstract visual thinking, as their flexibility is fundamentally limited by external tools. In this work, we introduce Monet, a training framework that enables multimodal large language models (MLLMs) to reason directly within the latent visual space by generating continuous embeddings that function as intermediate visual thoughts. We identify two core challenges in training MLLMs for latent visual reasoning: high computational cost in latent-vision alignment and insufficient supervision over latent embeddings, and address them with a three-stage distillation-based supervised fine-tuning (SFT) pipeline. We further reveal a limitation of applying GRPO to latent reasoning: it primarily enhances text-based reasoning rather than latent reasoning. To overcome this, we propose VLPO (Visual-latent Policy Optimization), a reinforcement learning method that explicitly incorporates latent embeddings into policy gradient updates. To support SFT, we construct Monet-SFT-125K, a high-quality text-image interleaved CoT dataset containing 125K real-world, chart, OCR, and geometry CoTs. Our model, Monet-7B, shows consistent gains across real-world perception and reasoning benchmarks and exhibits strong out-of-distribution generalization on challenging abstract visual reasoning tasks. We also empirically analyze the role of each training component and discuss our early unsuccessful attempts, providing insights for future developments in visual latent reasoning. Our model, data, and code are available at https://github.com/NOVAglow646/Monet.

cs.CV↗

arXiv2025

Leveraging Evidence-Guided LLMs to Enhance Trustworthy Depression Diagnosis

Large language models (LLMs) show promise in automating clinical diagnosis, yet their non-transparent decision-making and limited alignment with diagnostic standards hinder trust and clinical adoption. We address this challenge by proposing a two-stage diagnostic framework that enhances transparency, trustworthiness, and reliability. First, we introduce Evidence-Guided Diagnostic Reasoning (EGDR), which guides LLMs to generate structured diagnostic hypotheses by interleaving evidence extraction with logical reasoning grounded in DSM-5 criteria. Second, we propose a Diagnosis Confidence Scoring (DCS) module that evaluates the factual accuracy and logical consistency of generated diagnoses through two interpretable metrics: the Knowledge Attribution Score (KAS) and the Logic Consistency Score (LCS). Evaluated on the D4 dataset with pseudo-labels, EGDR outperforms direct in-context prompting and Chain-of-Thought (CoT) across five LLMs. For instance, on OpenBioLLM, EGDR improves accuracy from 0.31 (Direct) to 0.76 and increases DCS from 0.50 to 0.67. On MedLlama, DCS rises from 0.58 (CoT) to 0.77. Overall, EGDR yields up to +45% accuracy and +36% DCS gains over baseline methods, offering a clinically grounded, interpretable foundation for trustworthy AI-assisted diagnosis.

cs.AI↗

arXiv2025

An Augmentation Overlap Theory of Contrastive Learning

Recently, self-supervised contrastive learning has achieved great success on various tasks. However, its underlying working mechanism is yet unclear. In this paper, we first provide the tightest bounds based on the widely adopted assumption of conditional independence. Further, we relax the conditional independence assumption to a more practical assumption of augmentation overlap and derive the asymptotically closed bounds for the downstream performance. Our proposed augmentation overlap theory hinges on the insight that the support of different intra-class samples will become more overlapped under aggressive data augmentations, thus simply aligning the positive samples (augmented views of the same sample) could make contrastive learning cluster intra-class samples together. Moreover, from the newly derived augmentation overlap perspective, we develop an unsupervised metric for the representation evaluation of contrastive learning, which aligns well with the downstream performance almost without relying on additional modules. Code is available at https://github.com/PKU-ML/GARC.

cs.LG↗