arXiv subjects

Yin Zhang

Publications and source records attributed to Yin Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

JW-VL: A Vision-Language Model for Solar Physics

Vision-Language Models (VLMs) have achieved breakthrough progress in general knowledge domains, yet adaptation to specialized scientific fields remains challenging due to multimodal representation shifts and the limited integration of domain-specific knowledge. To address the limitations of general-purpose VLMs when applied to solar physics image recognition, analysis, and reasoning, we propose JinWu Vision-Language (JW-VL), a fine-tuned foundation model tailored for solar physics. The model integrates multi-wavelength observational data from both space-based and ground-based telescopes, encompassing representative spectral bands spanning the photosphere, chromosphere, and corona. Built upon a cross-modal alignment knowledge distillation framework, JW-VL learns a joint visual-semantic embedding that enables end-to-end modeling from raw solar observational data to downstream tasks, including solar image recognition, solar activity analysis via image-based question answering, and optical character recognition (OCR), while also supporting the construction of a multi-band, cross-instrument solar image benchmark dataset. Furthermore, as a demonstration of interdisciplinary applicability, we developed a "Daily Solar Activity Reports" agent comprising core modules for solar activity level assessment, significant active region characterization, magnetic field complexity analysis, potential space weather impact assessment, and identifying active regions for targeted observation. While JW-VL may not yet meet the rigorous, high-precision demands of operational solar physics, it bridges raw observations and diverse downstream tasks, establishing a valuable methodological framework for applying multimodal deep learning to the field.

astro-ph.SR↗

arXiv2026

Conversational Speech Reveals Structural Robustness Failures in SpeechLLM Backbones

LLMs serve as the backbone in SpeechLLMs, yet their behavior on spontaneous conversational input remains poorly understood. Conversational speech contains pervasive disfluencies -- interjections, edits, and parentheticals -- that are rare in the written corpora used for pre-training. Because gold disfluency removal is a deletion-only task, it serves as a controlled probe to determine whether a model performs faithful structural repair or biased reinterpretation. Using the DRES evaluation framework, we evaluate proprietary and open-source LLMs across architectures and scales. We show that model performance clusters into stable precision-recall regimes reflecting distinct editing policies. Notably, reasoning models systematically over-delete fluent content, revealing a bias toward semantic abstraction over structural fidelity. While fine-tuning achieves SOTA results, it harms generalization. Our findings demonstrate that robustness to speech is shaped by specific training objectives.

cs.CL↗

arXiv2026

GuardAlign: Test-time Safety Alignment in Multimodal Large Language Models

Large vision-language models (LVLMs) have achieved remarkable progress in vision-language reasoning tasks, yet ensuring their safety remains a critical challenge. Recent input-side defenses detect unsafe images with CLIP and prepend safety prefixes to prompts, but they still suffer from inaccurate detection in complex scenes and unstable safety signals during decoding. To address these issues, we propose GuardAlign, a training-free defense framework that integrates two strategies. First, OT-enhanced safety detection leverages optimal transport to measure distribution distances between image patches and unsafe semantics, enabling accurate identification of malicious regions without additional computational cost. Second, cross-modal attentive calibration strengthens the influence of safety prefixes by adaptively reallocating attention across layers, ensuring that safety signals remain consistently activated throughout generation. Extensive evaluations on six representative MLLMs demonstrate that GuardAlign reduces unsafe response rates by up to 39% on SPA-VL, while preserving utility, achieving an improvement on VQAv2 from 78.51% to 79.21%.

cs.CV↗

arXiv2026

Revisiting Weight Regularization for Low-Rank Continual Learning

Continual Learning (CL) with large-scale pre-trained models (PTMs) has recently gained wide attention, shifting the focus from training from scratch to continually adapting PTMs. This has given rise to a promising paradigm: parameter-efficient continual learning (PECL), where task interference is typically mitigated by assigning a task-specific module during training, such as low-rank adapters. However, weight regularization techniques, such as Elastic Weight Consolidation (EWC)-a key strategy in CL-remain underexplored in this new paradigm. In this paper, we revisit weight regularization in low-rank CL as a new perspective for mitigating task interference in PECL. Unlike existing low-rank CL methods, we mitigate task interference by regularizing a shared low-rank update through EWC, thereby keeping the storage requirement and inference costs constant regardless of the number of tasks. Our proposed method EWC-LoRA leverages a low-rank representation to estimate parameter importance over the full-dimensional space. This design offers a practical, computational- and memory-efficient solution for CL with PTMs, and provides insights that may inform the broader application of regularization techniques within PECL. Extensive experiments on various benchmarks demonstrate the effectiveness of EWC-LoRA, achieving a stability-plasticity trade-off superior to existing low-rank CL approaches. These results indicate that, even under low-rank parameterizations, weight regularization remains an effective mechanism for mitigating task interference. Code is available at: https://github.com/yaoyz96/low-rank-cl.

cs.LG↗

arXiv2026

VideoPro: Adaptive Program Reasoning for Long Video Understanding

Large language models (LLMs) have shown promise in generating program workflows for visual tasks. However, previous approaches often rely on closed-source models, lack systematic reasoning, and struggle with long-form video question answering (videoQA). To address these challenges, we introduce the FS-VisPR framework, an adaptive visual program reasoning approach that balances fast reasoning for simple queries with slow reasoning for difficult ones. First, we design efficient visual modules (e.g., key clip retrieval and subtitle retrieval) to support long-form video tasks. Then, we construct a diverse and high-quality fast-slow reasoning dataset with a strong LLM to align open-source language models' ability to generate visual program workflows as FS-LLM. Next, we design a fast-slow reasoning framework with FS-LLM: Simple queries are directly solved by VideoLLMs, while difficult ones invoke visual program reasoning, motivated by human-like reasoning processes. During this process, low-confidence fast-thinking answers will trigger a second-stage slow-reasoning process, and a fallback mechanism to fast reasoning is activated if the program execution fails. Moreover, we improve visual programs through parameter search during both training and inference. By adjusting the parameters of the visual modules within the program, multiple variants are generated: during training, programs that yield correct answers are selected, while during inference, the program with the highest confidence result is applied. Experiments show that FS-VisPR improves both efficiency and reliability in visual program workflows. It achieves 50.4% accuracy on LVBench, surpassing GPT-4o, matching the performance of Qwen2.5VL-72B on VideoMME.

cs.CV↗

arXiv2026

ShortCoder: Knowledge-Augmented Syntax Optimization for Token-Efficient Code Generation

Code generation tasks aim to automate the conversion of user requirements into executable code, significantly reducing manual development efforts and enhancing software productivity. The emergence of large language models (LLMs) has significantly advanced code generation, though their efficiency is still impacted by certain inherent architectural constraints. Each token generation necessitates a complete inference pass, requiring persistent retention of contextual information in memory and escalating resource consumption. While existing research prioritizes inference-phase optimizations such as prompt compression and model quantization, the generation phase remains underexplored. To tackle these challenges, we propose a knowledge-infused framework named ShortCoder, which optimizes code generation efficiency while preserving semantic equivalence and readability. In particular, we introduce: (1) ten syntax-level simplification rules for Python, derived from AST-preserving transformations, achieving 18.1% token reduction without functional compromise; (2) a hybrid data synthesis pipeline integrating rule-based rewriting with LLM-guided refinement, producing ShorterCodeBench, a corpus of validated tuples of original code and simplified code with semantic consistency; (3) a fine-tuning strategy that injects conciseness awareness into the base LLMs. Extensive experimental results demonstrate that ShortCoder consistently outperforms state-of-the-art methods on HumanEval, achieving an improvement of 18.1%-37.8% in generation efficiency over previous methods while ensuring the performance of code generation.

cs.SE↗

arXiv2026

A microscopic origin for the breakdown of the Stokes Einstein relation in ion transport

Ion transport underlies the operation of biological ion channels and governs the performance of electrochemical energy-storage devices. A long-standing anomaly is that smaller alkali metal ions, such as Li$^+$, migrate more slowly in water than larger ions, in apparent violation of the Stokes-Einstein relation. This breakdown is conventionally attributed to dielectric friction, a collective drag force arising from electrostatic interactions between a drifting ion and its surrounding solvent. Here, combining nanopore transport measurements over electric fields spanning several orders of magnitude with molecular dynamics simulations, we show that the time-averaged electrostatic force on a migrating ion is not a drag force but a net driving force. By contrasting charged ions with neutral particles, we reveal that ionic charge introduces additional Lorentzian peaks in the frequency-dependent friction coefficient. These peaks originate predominantly from short-range Lennard-Jones (LJ) interactions within the first hydration layer and represent additional channels for energy dissipation, strongest for Li$^+$ and progressively weaker for Na$^+$ and K$^+$. Our results demonstrate that electrostatic interactions primarily act to tighten the local hydration structure, thereby amplifying short-range LJ interactions rather than directly opposing ion motion. This microscopic mechanism provides a unified physical explanation for the breakdown of the Stokes-Einstein relation in aqueous ion transport.

cond-mat.soft↗

arXiv2026

Observability-Enhanced Target Motion Estimation via Bearing-Box: Theory and MAV Applications

Monocular vision-based target motion estimation is a fundamental challenge in numerous applications. This work introduces a novel bearing-box approach that fully leverages modern 3D detection measurements that are widely available nowadays but have not been well explored for motion estimation so far. Unlike existing methods that rely on restrictive assumptions such as isotropic target shape and lateral motion, our bearing-box estimator can estimate both the target's motion and its physical size without these assumptions by exploiting the information buried in a 3D bounding box. When applied to multi-rotor micro aerial vehicles (MAVs), the estimator yields an interesting advantage: it further removes the need for higher-order motion assumptions by exploiting the unique coupling between MAV's acceleration and thrust. This is particularly significant, as higher-order motion assumptions are widely believed to be necessary in state-of-the-art bearing-based estimators. We support our claims with rigorous observability analyses and extensive experimental validation, demonstrating the estimator's superior performance in real-world scenarios.

cs.RO↗

arXiv2025

Training Report of TeleChat3-MoE

TeleChat3-MoE is the latest series of TeleChat large language models, featuring a Mixture-of-Experts (MoE) architecture with parameter counts ranging from 105 billion to over one trillion,trained end-to-end on Ascend NPU cluster. This technical report mainly presents the underlying training infrastructure that enables reliable and efficient scaling to frontier model sizes. We detail systematic methodologies for operator-level and end-to-end numerical accuracy verification, ensuring consistency across hardware platforms and distributed parallelism strategies. Furthermore, we introduce a suite of performance optimizations, including interleaved pipeline scheduling, attention-aware data scheduling for long-sequence training,hierarchical and overlapped communication for expert parallelism, and DVM-based operator fusion. A systematic parallelization framework, leveraging analytical estimation and integer linear programming, is also proposed to optimize multi-dimensional parallelism configurations. Additionally, we present methodological approaches to cluster-level optimizations, addressing host- and device-bound bottlenecks during large-scale training tasks. These infrastructure advancements yield significant throughput improvements and near-linear scaling on clusters comprising thousands of devices, providing a robust foundation for large-scale language model development on hardware ecosystems.

cs.CL↗

arXiv2025

Causal-Tune: Mining Causal Factors from Vision Foundation Models for Domain Generalized Semantic Segmentation

Fine-tuning Vision Foundation Models (VFMs) with a small number of parameters has shown remarkable performance in Domain Generalized Semantic Segmentation (DGSS). Most existing works either train lightweight adapters or refine intermediate features to achieve better generalization on unseen domains. However, they both overlook the fact that long-term pre-trained VFMs often exhibit artifacts, which hinder the utilization of valuable representations and ultimately degrade DGSS performance. Inspired by causal mechanisms, we observe that these artifacts are associated with non-causal factors, which usually reside in the low- and high-frequency components of the VFM spectrum. In this paper, we explicitly examine the causal and non-causal factors of features within VFMs for DGSS, and propose a simple yet effective method to identify and disentangle them, enabling more robust domain generalization. Specifically, we propose Causal-Tune, a novel fine-tuning strategy designed to extract causal factors and suppress non-causal ones from the features of VFMs. First, we extract the frequency spectrum of features from each layer using the Discrete Cosine Transform (DCT). A Gaussian band-pass filter is then applied to separate the spectrum into causal and non-causal components. To further refine the causal components, we introduce a set of causal-aware learnable tokens that operate in the frequency domain, while the non-causal components are discarded. Finally, refined features are transformed back into the spatial domain via inverse DCT and passed to the next layer. Extensive experiments conducted on various cross-domain tasks demonstrate the effectiveness of Causal-Tune. In particular, our method achieves superior performance under adverse weather conditions, improving +4.8% mIoU over the baseline in snow conditions.

cs.CV↗

arXiv2025

2K-Characters-10K-Stories: A Quality-Gated Stylized Narrative Dataset with Disentangled Control and Sequence Consistency

Sequential identity consistency under precise transient attribute control remains a long-standing challenge in controllable visual storytelling. Existing datasets lack sufficient fidelity and fail to disentangle stable identities from transient attributes, limiting structured control over pose, expression, and scene composition and thus constraining reliable sequential synthesis. To address this gap, we introduce \textbf{2K-Characters-10K-Stories}, a multi-modal stylized narrative dataset of \textbf{2{,}000} uniquely stylized characters appearing across \textbf{10{,}000} illustration stories. It is the first dataset that pairs large-scale unique identities with explicit, decoupled control signals for sequential identity consistency. We introduce a \textbf{Human-in-the-Loop pipeline (HiL)} that leverages expert-verified character templates and LLM-guided narrative planning to generate highly-aligned structured data. A \textbf{decoupled control} scheme separates persistent identity from transient attributes -- pose and expression -- while a \textbf{Quality-Gated loop} integrating MMLM evaluation, Auto-Prompt Tuning, and Local Image Editing enforces pixel-level consistency. Extensive experiments demonstrate that models fine-tuned on our dataset achieves performance comparable to closed-source models in generating visual narratives.

cs.CV↗

arXiv2025

Multiwavelength Observations for a Double-decker Filament Channel in AR 13102

We present the observational evidence of the existence of a double-decker filament channel (FC) by using observations in extreme ultraviolet and Hα wavelengths. For both FCs, the east foot-point roots in the active region (AR), while the west one roots in the remote quiet region. The bottom FC (FC1) appears as intermittent filaments. Within the AR, the FC1 appears as an S-shaped filament (F1), which consisted of two J-shaped filaments (F1S/F1N for the south/north one). For the upper one (FC2), only the east part is filled with dark plasma and visible as a small filament (F2). Its east foot-point roots around the junction of F1S and F1N. Initially, due to the recurrent reconnections, F1N and F1S link to each other and form a new filament (F3) thread by thread. Meanwhile, the heated plasma, which appears as brightening features, flows from the east foot-point of F2 to the west, and becomes invisible about 1.1$\times$10^{5} km away. The failed eruption of F1S is triggered by the reconnection, which appears as the brightening threads changing their configuration from crossed to quasiparallel in between the F1S and F3, and is confined by the upper magnetic field. Associated with the eruption, the distant invisible plasma becomes visible as a brightening feature. It continuously flows to the remote foot-point, and becomes invisible before reaching it. The brightening plasma flow outlines the skeleton of FC2 gradually. The observations show the existence of a double-decker FC, as a magnetic structure, before they appear as a brightening/dark feature when fully filled with hot/cool plasma.

astro-ph.SR↗

arXiv2025

JW-Flare: Accurate Solar Flare Forecasting Method Based on Multimodal Large Language Models

Solar flares, the most powerful explosive phenomena in the solar system, may pose significant hazards to spaceborne satellites and ground-based infrastructure. Despite decades of intensive research, reliable flare prediction remains a challenging task. Large Language Models, as a milestone in artificial intelligence, exhibit exceptional general knowledge and next-token prediction capabilities. Here we introduce JW-Flare, the first Multimodal Large Language Models (MLLMs) explicitly trained for solar flare forecasting through fine-tuning on textual physic parameters of solar active regions and magnetic field images. This method demonstrates state-of-the-art (SOTA) performance for large flares prediction on the test dataset. It effectively identifies all 79 X-class flares from 18,949 test samples, yielding a True Skill Statistic (TSS) of 0.95 and a True Positive Rate (TPR) of 1.00, outperforming traditional predictive models. We further investigate the capability origins of JW-Flare through explainability experiments, revealing that solar physics knowledge acquired during pre-training contributes to flare forecasting performance. Additionally, we evaluate models of different parameter scales, confirming the Scaling_Law of Large Language Models in domain-specific applications, such as solar physics. This study marks a substantial advance in both the scale and accuracy of solar flare forecasting and opens a promising avenue for AI-driven methodologies in broader scientific domains.

astro-ph.SR↗

arXiv2025

Weed Out, Then Harvest: Dual Low-Rank Adaptation is an Effective Noisy Label Detector for Noise-Robust Learning

Parameter-efficient fine-tuning (PEFT) large language models (LLMs) have shown impressive performance in various downstream tasks. However, in many real-world scenarios, the collected training data inevitably contains noisy labels. To learn from noisy labels, most solutions select samples with small losses for model training. However, the selected samples, in turn, impact the loss computation in the next iteration. An inaccurate initial selection can create a vicious cycle, leading to suboptimal performance. To break this cycle, we propose Delora, a novel framework that decouples the sample selection from model training. For sample selection, Delora establishes a noisy label detector by introducing clean and noisy LoRA. Benefiting from the memory effect, the clean LoRA is encouraged to memorize clean data, while the noisy LoRA is constrained to memorize mislabeled data, which serves as a learnable threshold for selecting clean and noisy samples. For model training, Delora can use carefully selected samples to fine-tune language models seamlessly. Experimental results on synthetic and real-world noisy datasets demonstrate the effectiveness of Delora in noisy label detection and text classification.

cs.CL↗

arXiv2025

QuadEnhancer: Leveraging Quadratic Transformations to Enhance Deep Neural Networks

The combination of linear transformations and non-linear activation functions forms the foundation of most modern deep neural networks, enabling them to approximate highly complex functions. This paper explores the introduction of quadratic transformations to further increase nonlinearity in neural networks, with the aim of enhancing the performance of existing architectures. To reduce parameter complexity and computational complexity, we propose a lightweight quadratic enhancer that uses low-rankness, weight sharing, and sparsification techniques. For a fixed architecture, the proposed approach introduces quadratic interactions between features at every layer, while only adding negligible amounts of additional model parameters and forward computations. We conduct a set of proof-of-concept experiments for the proposed method across three tasks: image classification, text classification, and fine-tuning large-language models. In all tasks, the proposed approach demonstrates clear and substantial performance gains.

cs.LG↗

arXiv2025

Z-Scores: A Metric for Linguistically Assessing Disfluency Removal

Evaluating disfluency removal in speech requires more than aggregate token-level scores. Traditional word-based metrics such as precision, recall, and F1 (E-Scores) capture overall performance but cannot reveal why models succeed or fail. We introduce Z-Scores, a span-level linguistically-grounded evaluation metric that categorizes system behavior across distinct disfluency types (EDITED, INTJ, PRN). Our deterministic alignment module enables robust mapping between generated text and disfluent transcripts, allowing Z-Scores to expose systematic weaknesses that word-level metrics obscure. By providing category-specific diagnostics, Z-Scores enable researchers to identify model failure modes and design targeted interventions -- such as tailored prompts or data augmentation -- yielding measurable performance improvements. A case study with LLMs shows that Z-Scores uncover challenges with INTJ and PRN disfluencies hidden in aggregate F1, directly informing model refinement strategies.

cs.CL↗

arXiv2025

PIN: A Knowledge-Intensive Dataset for Paired and Interleaved Multimodal Documents

Recent advancements in large multimodal models (LMMs) have leveraged extensive multimodal datasets to enhance capabilities in complex knowledge-driven tasks. However, persistent challenges in perceptual and reasoning errors limit their efficacy, particularly in interpreting intricate visual data and deducing multimodal relationships. To address these issues, we introduce PIN (Paired and INterleaved multimodal documents), a novel data format designed to foster a deeper integration of visual and textual knowledge. The PIN format uniquely combines semantically rich Markdown files, which preserve fine-grained textual structures, with holistic overall images that capture the complete document layout. Following this format, we construct and release two large-scale, open-source datasets: PIN-200M (~200 million documents) and PIN-14M (~14 million), compiled from diverse web and scientific sources in both English and Chinese. To maximize usability, we provide detailed statistical analyses and equip the datasets with quality signals, enabling researchers to easily filter and select data for specific tasks. Our work provides the community with a versatile data format and substantial resources, offering a foundation for new research in pre-training strategies and the development of more powerful knowledge-intensive LMMs.

cs.AI↗

arXiv2025

Casual3DHDR: Deblurring High Dynamic Range 3D Gaussian Splatting from Casually Captured Videos

Photo-realistic novel view synthesis from multi-view images, such as neural radiance field (NeRF) and 3D Gaussian Splatting (3DGS), has gained significant attention for its superior performance. However, most existing methods rely on low dynamic range (LDR) images, limiting their ability to capture detailed scenes in high-contrast environments. While some prior works address high dynamic range (HDR) scene reconstruction, they typically require multi-view sharp images with varying exposure times captured at fixed camera positions, which is time-consuming and impractical. To make data acquisition more flexible, we propose \textbf{Casual3DHDR}, a robust one-stage method that reconstructs 3D HDR scenes from casually-captured auto-exposure (AE) videos, even under severe motion blur and unknown, varying exposure times. Our approach integrates a continuous-time camera trajectory into a unified physical imaging model, jointly optimizing exposure times, camera trajectory, and the camera response function (CRF). Extensive experiments on synthetic and real-world datasets demonstrate that \textbf{Casual3DHDR} outperforms existing methods in robustness and rendering quality. Our source code and dataset will be available at https://lingzhezhao.github.io/CasualHDRSplat/

cs.CV↗