arXiv subjects

Yan Li

Publications and source records attributed to Yan Li.

At least 199 records · Page 11Linked to original sources

arXiv2025

Exploiting Block Coordinate Descent for Cost-Effective LLM Model Training

Training large language models typically demands extensive GPU memory and substantial financial investment, which poses a barrier for many small- to medium-sized teams. In this paper, we propose a full-parameter pre-training and fine-tuning framework based on block coordinate descent (BCD), enhanced with engineering optimizations, to enable efficient training of large-scale models on cost-effective RTX 4090, A100 and A800 GPU clusters. Under identical hardware configurations, we reduce the training cost of a 7B model to 33% on A100/A800 and only 2.6% on RTX 4090, compared to standard full-parameter training. It also enables large models previously restricted to A100 clusters to be trained on RTX 4090 without degrading performance. BCD achieves comparable or better accuracy than full-parameter and fine-tuning methods at most cases, with lower GPU consumption and improved hardware utilization.

cs.LG↗

arXiv2025

Convergence of the adaptive finite element discretization based parallel orbital-updating method for eigenvalue problems

It is significant and challenging to solve eigenvalue problems of partial differential operators when many highly accurate eigenpair approximations are required. The adaptive finite element discretization based parallel orbital-updating method, which can significantly reduce the computational cost and enhance the parallel scalability, has been shown to be efficient in electronic structure calculations. However, there is no any mathematical justification for this method in literature. In this paper, we will show the convergence of the method for clustered eigenvalue problems of linear partial differential operators.

math.NA↗

arXiv2025

Electrically Reconfigurable Arbitrary Splitting-Ratio Optical Splitter Based on Low-Loss Sb2Se3

Reconfigurable beam splitters capable of being arbitrarily programmed for the power splitting ratios are vital for the adaptive optical networks and photonic computing. Conventional mechanisms such as thermo-optic, free-carrier, or mechanical tuning are usually volatile and require continuous power, limiting their suitability for low-frequency and low power-consumption programmable operations. Here, we experimentally demonstrate an electrically reconfigurable beam splitter based on the low-loss phase-change material Sb2Se3, enabling multi-level and arbitrary splitting-ratio (SR) control. By locally triggering phase transitions in the coupling region with integrated micro-electrodes, we exploit the high refractive-index contrast between different phases and negligible absorption in the near-infrared wavelength of Sb2Se3 to precisely tune the coupling strength with non-volatile retention. 8-level of power splitting states is achieved within a compact footprint of ~14.5-μm in the experiments, with insertion loss is ~1 dB across 1515-1550 nm and near-zero static power. Combining the advantages of compactness, broad bandwidth, low loss, non-volatility, and multi-level control experimentally, this device provides a universal building block for scalable, energy-efficient reconfigurable photonic circuits, with great prospects in optical computing and intelligent communication systems.

physics.optics↗

arXiv2025

Phase Behavior and Ion Transport in Lithium-Niobium-Tantalum Oxide Alloys

Lithium niobate-tantalate mixtures have garnered considerable interest for their ability to merge the desirable properties of both end members, enabling diverse high-value applications, such as high-performance faradaic capacitors, non-linear optics, and protective coatings in rechargeable batteries. While numerous studies on the application of $\mathrm{LiNb_xTa_{1-x}O_3}$ exist, the phase behavior and properties of $\mathrm{Li_3Nb_xTa_{1-x}O_4}$ remain largely unexplored. In this work, we employ a multiscale approach that encompasses first-principles phonon calculations, cluster expansion, and Monte Carlo simulations to derive the temperature-composition phase diagram for $\mathrm{Li_3Nb_xTa_{1-x}O_4}$. Our findings reveal the critical role of vibrational entropy in accurately predicting phase stability, which promotes the solubility of Nb in $\mathrm{Li_3TaO_4}$ while suppressing the miscibility of Ta in $\mathrm{Li_3NbO_4}$. Additionally, we demonstrate that Nb/Ta mixing offers a promising avenue for tailoring the Li-ion conductivities of $\mathrm{Li_3Nb_xTa_{1-x}O_4}$. On the technical side, we demonstrated the importance of including vibrational entropy effects explicitly in Monte Carlo simulations dealing with multicomponent systems, beyond simple binary mixtures. On the application side, this study provides fundamental insights into the phase behavior and Li-ion transport properties of $\mathrm{Li_3Nb_xTa_{1-x}O_4}$, paving the way for its potential applications in energy storage and other fields.

cond-mat.mtrl-sci↗

arXiv2025

R-Net: A Reliable and Resource-Efficient CNN for Colorectal Cancer Detection with XAI Integration

State-of-the-art (SOTA) Convolutional Neural Networks (CNNs) are criticized for their extensive computational power, long training times, and large datasets. To overcome this limitation, we propose a reasonable network (R-Net), a lightweight CNN only to detect and classify colorectal cancer (CRC) using the Enteroscope Biopsy Histopathological Hematoxylin and Eosin Image Dataset (EBHI). Furthermore, six SOTA CNNs, including Multipath-based CNNs (DenseNet121, ResNet50), Depth-based CNNs (InceptionV3), width-based multi-connection CNNs (Xception), depth-wise separable convolutions (MobileNetV2), spatial exploitation-based CNNs (VGG16), Transfer learning, and two ensemble models are also tested on the same dataset. The ensemble models are a multipath-depth-width combination (DenseNet121-InceptionV3-Xception) and a multipath-depth-spatial combination (ResNet18-InceptionV3-VGG16). However, the proposed R-Net lightweight achieved 99.37% accuracy, outperforming MobileNet (95.83%) and ResNet50 (96.94%). Most importantly, to understand the decision-making of R-Net, Explainable AI such as SHAP, LIME, and Grad-CAM are integrated to visualize which parts of the EBHI image contribute to the detection and classification process of R-Net. The main novelty of this research lies in building a reliable, lightweight CNN R-Net that requires fewer computing resources yet maintains strong prediction results. SOTA CNNs, transfer learning, and ensemble models also extend our knowledge on CRC classification and detection. XAI functionality and the impact of pixel intensity on correct and incorrect classification images are also some novelties in CRC detection and classification.

q-bio.TO↗

arXiv2025

On the Equivalence of Koopman Eigenfunctions and Commuting Symmetries

The Koopman operator framework offers a way to represent a nonlinear system as a linear one. The key to this simplification lies in the identification of eigenfunctions. While various data-driven algorithms have been developed for this problem, a theoretical characterization of Koopman eigenfunctions from geometric properties of the flow is still missing. This paper provides such a characterization by establishing an equivalence between a set of Koopman eigenfunctions and a set of commuting symmetries -- both assumed to span the tangent spaces at every point on a simply connected open set. Based on this equivalence, we derive an explicit formula for the principal Koopman eigenfunctions and prove its uniform convergence on the region of attraction of a locally asymptotically stable equilibrium point, thereby offering a constructive method for computing Koopman eigenfunctions.

eess.SY↗

arXiv2025

Kwai Keye-VL 1.5 Technical Report

In recent years, the development of Large Language Models (LLMs) has significantly advanced, extending their capabilities to multimodal tasks through Multimodal Large Language Models (MLLMs). However, video understanding remains a challenging area due to the dynamic and information-dense nature of videos. Existing models struggle with the trade-off between spatial resolution and temporal coverage when processing video content. We present Keye-VL-1.5, which addresses fundamental challenges in video comprehension through three key innovations. First, we introduce a novel Slow-Fast video encoding strategy that dynamically allocates computational resources based on inter-frame similarity, processing key frames with significant visual changes at higher resolution (Slow pathway) while handling relatively static frames with increased temporal coverage at lower resolution (Fast pathway). Second, we implement a progressive four-stage pre-training methodology that systematically extends the model's context length from 8K to 128K tokens, enabling processing of longer videos and more complex visual content. Third, we develop a comprehensive post-training pipeline focusing on reasoning enhancement and human preference alignment, incorporating a 5-step chain-of-thought data construction process, iterative GSPO-based reinforcement learning with progressive prompt hinting for difficult cases, and alignment training. Through extensive evaluation on public benchmarks and rigorous internal human assessment, Keye-VL-1.5 demonstrates significant improvements over existing models, particularly excelling in video understanding tasks while maintaining competitive performance on general multimodal benchmarks.

cs.CV↗

arXiv2025

Meridional Circulation II: A Unified Mechanism for Lithium Depletion in Solar Analogs and the Lithium Dip in Mid-F Cluster Stars

The behavior of lithium (Li) in Population I main sequence stars challenges standard stellar theory. Two phenomena stand out: the solar Li problem which extends to Li depletion in solar analogs and the Li dip observed in mid-F stars within open clusters. Building on the meridional circulation-driven radial mixing framework previously developed to explain Li-enriched red clump stars, we explore its relevance to Li depletion on the main sequence. First, our models reproduce the observed $A(\text{Li})$-Age correlation in solar analogs. Through detailed isochrone analysis, we find good agreement between the simulated and observed $A(\text{Li})$-$T_{\text{eff}}$ relationships within the solar analog parameter space. However, the predicted solar Li abundance ($\sim 1.5\,\text{dex}$) is still higher than current solar measurements. Second, our models partially explain the Li dip phenomenon in mid-F cluster stars. The models accurately reproduce Li distributions on the cool side of the Li dip in most clusters and capture the Li behaviors on the hot side observed in systems like the Hyades. However, we identify limitations in the models' ability to fully reproduce the dip morphology, particularly due to the rotation velocity distribution of sample stars in this temperature zone.

astro-ph.SR↗

arXiv2025

UniAPO: Unified Multimodal Automated Prompt Optimization

Prompting is fundamental to unlocking the full potential of large language models. To automate and enhance this process, automatic prompt optimization (APO) has been developed, demonstrating effectiveness primarily in text-only input scenarios. However, extending existing APO methods to multimodal tasks, such as video-language generation introduces two core challenges: (i) visual token inflation, where long visual token sequences restrict context capacity and result in insufficient feedback signals; (ii) a lack of process-level supervision, as existing methods focus on outcome-level supervision and overlook intermediate supervision, limiting prompt optimization. We present UniAPO: Unified Multimodal Automated Prompt Optimization, the first framework tailored for multimodal APO. UniAPO adopts an EM-inspired optimization process that decouples feedback modeling and prompt refinement, making the optimization more stable and goal-driven. To further address the aforementioned challenges, we introduce a short-long term memory mechanism: historical feedback mitigates context limitations, while historical prompts provide directional guidance for effective prompt optimization. UniAPO achieves consistent gains across text, image, and video benchmarks, establishing a unified framework for efficient and transferable prompt optimization.

cs.CV↗

arXiv2025

ChuLo: Chunk-Level Key Information Representation for Long Document Understanding

Transformer-based models have achieved remarkable success in various Natural Language Processing (NLP) tasks, yet their ability to handle long documents is constrained by computational limitations. Traditional approaches, such as truncating inputs, sparse self-attention, and chunking, attempt to mitigate these issues, but they often lead to information loss and hinder the model's ability to capture long-range dependencies. In this paper, we introduce ChuLo, a novel chunk representation method for long document understanding that addresses these limitations. Our ChuLo groups input tokens using unsupervised keyphrase extraction, emphasizing semantically important keyphrase based chunks to retain core document content while reducing input length. This approach minimizes information loss and improves the efficiency of Transformer-based models. Preserving all tokens in long document understanding, especially token classification tasks, is important to ensure that fine-grained annotations, which depend on the entire sequence context, are not lost. We evaluate our method on multiple long document classification tasks and long document token classification tasks, demonstrating its effectiveness through comprehensive qualitative and quantitative analysis. Our implementation is open-sourced on https://github.com/adlnlp/Chulo.

cs.CL↗

arXiv2025

MUSE: Multi-Subject Unified Synthesis via Explicit Layout Semantic Expansion

Existing text-to-image diffusion models have demonstrated remarkable capabilities in generating high-quality images guided by textual prompts. However, achieving multi-subject compositional synthesis with precise spatial control remains a significant challenge. In this work, we address the task of layout-controllable multi-subject synthesis (LMS), which requires both faithful reconstruction of reference subjects and their accurate placement in specified regions within a unified image. While recent advancements have separately improved layout control and subject synthesis, existing approaches struggle to simultaneously satisfy the dual requirements of spatial precision and identity preservation in this composite task. To bridge this gap, we propose MUSE, a unified synthesis framework that employs concatenated cross-attention (CCA) to seamlessly integrate layout specifications with textual guidance through explicit semantic space expansion. The proposed CCA mechanism enables bidirectional modality alignment between spatial constraints and textual descriptions without interference. Furthermore, we design a progressive two-stage training strategy that decomposes the LMS task into learnable sub-objectives for effective optimization. Extensive experiments demonstrate that MUSE achieves zero-shot end-to-end generation with superior spatial accuracy and identity consistency compared to existing solutions, advancing the frontier of controllable image synthesis. Our code and model are available at https://github.com/pf0607/MUSE.

cs.CV↗

arXiv2025

Co-Paced Learning Strategy Based on Confidence for Flying Bird Object Detection Model Training

The flying bird objects captured by surveillance cameras exhibit varying levels of recognition difficulty due to factors such as their varying sizes or degrees of similarity to the background. To alleviate the negative impact of hard samples on training the Flying Bird Object Detection (FBOD) model for surveillance videos, we propose the Co-Paced Learning strategy Based on Confidence (CPL-BC) and apply it to the training process of the FBOD model. This strategy involves maintaining two models with identical structures but different initial parameter configurations that collaborate with each other to select easy samples for training, where the prediction confidence exceeds a set threshold. As training progresses, the strategy gradually lowers the threshold, thereby gradually enhancing the model's ability to recognize objects, from easier to more hard ones. Prior to applying CPL-BC, we pre-trained the two FBOD models to equip them with the capability to assess the difficulty of flying bird object samples. Experimental results on two different datasets of flying bird objects in surveillance videos demonstrate that, compared to other model learning strategies, CPL-BC significantly improves detection accuracy, thereby verifying the method's effectiveness and advancement.

cs.CV↗

arXiv2025

A Steel Surface Defect Detection Method Based on Lightweight Convolution Optimization

Surface defect detection of steel, especially the recognition of multi-scale defects, has always been a major challenge in industrial manufacturing. Steel surfaces not only have defects of various sizes and shapes, which limit the accuracy of traditional image processing and detection methods in complex environments. However, traditional defect detection methods face issues of insufficient accuracy and high miss-detection rates when dealing with small target defects. To address this issue, this study proposes a detection framework based on deep learning, specifically YOLOv9s, combined with the C3Ghost module, SCConv module, and CARAFE upsampling operator, to improve detection accuracy and model performance. First, the SCConv module is used to reduce feature redundancy and optimize feature representation by reconstructing the spatial and channel dimensions. Second, the C3Ghost module is introduced to enhance the model's feature extraction ability by reducing redundant computations and parameter volume, thereby improving model efficiency. Finally, the CARAFE upsampling operator, which can more finely reorganize feature maps in a content-aware manner, optimizes the upsampling process and ensures detailed restoration of high-resolution defect regions. Experimental results demonstrate that the proposed model achieves higher accuracy and robustness in steel surface defect detection tasks compared to other methods, effectively addressing defect detection problems.

eess.IV↗

arXiv2025

Multi-View Clustering Meets Heterogenous Data: A Fusion Regularized Method

Multi-view clustering leverages consistent and complementary information across multiple views to provide more comprehensive insights than single-view analysis. However, the heterogeneity and redundancy of multi-view data pose significant challenges to the existing clustering techniques. To tackle these challenges effectively, this paper proposes a novel multi-view fusion regularized clustering method with adaptive group sparsity, enabling discriminative clustering while capturing informative features. Technically, for heterogeneous multi-view data with mixed-type feature sets, different losses or divergence metrics are considered with a joint fusion penalty to obtain consistent cluster structures. Moreover, the non-convex group sparsity consisting of inter-group sparsity and intra-group sparsity is utilized to eliminate redundant features, thereby enhancing the robustness. Furthermore, we develop an effective alternating direction method of multipliers (ADMM), where all subproblems can be solved in closed form. Extensive numerical experiments on real data validate the superior performance of our presented method in clustering accuracy and feature selection.

math.OC↗

arXiv2025

Embodied Intelligence for 3D Understanding: A Survey on 3D Scene Question Answering

3D Scene Question Answering (3D SQA) represents an interdisciplinary task that integrates 3D visual perception and natural language processing, empowering intelligent agents to comprehend and interact with complex 3D environments. Recent advances in large multimodal modelling have driven the creation of diverse datasets and spurred the development of instruction-tuning and zero-shot methods for 3D SQA. However, this rapid progress introduces challenges, particularly in achieving unified analysis and comparison across datasets and baselines. In this survey, we provide the first comprehensive and systematic review of 3D SQA. We organize existing work from three perspectives: datasets, methodologies, and evaluation metrics. Beyond basic categorization, we identify shared architectural patterns across methods. Our survey further synthesizes core limitations and discusses how current trends, such as instruction tuning, multimodal alignment, and zero-shot, can shape future developments. Finally, we propose a range of promising research directions covering dataset construction, task generalization, interaction modeling, and unified evaluation protocols. This work aims to serve as a foundation for future research and foster progress toward more generalizable and intelligent 3D SQA systems.

cs.CV↗

arXiv2025

Gradient-Boosted Pseudo-Weighting: Methods for Population Inference from Nonprobability samples

Nonprobability samples have rapidly emerged to address time-sensitive priority topics in a variety of fields. While these data are timely, they are prone to selection bias. To mitigate selection bias, a large number of survey research literature has explored the use of propensity score (PS) adjustment methods to enhance population representativeness of nonprobability samples, using probability-based survey samples as external references. A recent advancement, the 2-step PS-based pseudo-weighting adjustment method (2PS, Li 2024), has been shown to improve upon recent developments with respect to mean squared error. However, the effectiveness of these methods in reducing bias critically depends on the ability of the underlying propensity model to accurately reflect the true selection process, which is challenging with parametric regression. In this study, we propose a set of pseudo-weight construction methods, which utilize gradient boosting methods (GBM) to estimate PSs in 2PS to construct pseudo-weights, offering greater flexibility compared to logistic regression-based methods. We compare the proposed GBM-based pseudo-weights with existing methods, including 2PS. The population mean estimators are evaluated via Monte Carlo simulation studies. We also evaluated prevalence of various health outcomes, including 15-year mortality, using 1988 ~ 1994 NHANES III as a nonprobability sample and the 1994 NHIS as the reference survey.

stat.ME↗

arXiv2025

Principle-Guided Verilog Optimization: IP-Safe Knowledge Transfer via Local-Cloud Collaboration

Recent years have witnessed growing interest in adopting large language models (LLMs) for Register Transfer Level (RTL) code optimization. While powerful cloud-based LLMs offer superior optimization capabilities, they pose unacceptable intellectual property (IP) leakage risks when processing proprietary hardware designs. In this paper, we propose a new scenario where Verilog code must be optimized for specific attributes without leaking sensitive IP information. We introduce the first IP-preserving edge-cloud collaborative framework that leverages the benefits of both paradigms. Our approach employs local small LLMs (e.g., Qwen-2.5-Coder-7B) to perform secure comparative analysis between paired high-quality target designs and novice draft codes, yielding general design principles that summarize key insights for improvements. These principles are then used to query stronger cloud LLMs (e.g., Deepseek-V3) for targeted code improvement, ensuring that only abstracted and IP-safe guidance reaches external services. Our experimental results demonstrate that the framework achieves significantly higher optimization success rates compared to baseline methods. For example, combining Qwen-2.5-Coder-7B and Deepseek-V3 achieves a 66.67\% optimization success rate for power utilization, outperforming Deepseek-V3 alone (49.81\%) and even commercial models like GPT-4o (55.81\%). Further investigation of local and cloud LLM combinations reveals that different model pairings exhibit varying strengths for specific optimization objectives, with interesting trends emerging when varying the number of comparative code pairs. Our work establishes a new paradigm for secure hardware design optimization that balances performance gains with IP protection.

cs.CR↗

arXiv2025

Effect of Matter Accretion on Lithium Enhancement of Giants

A subset of low-mass giants ($<2.2\,M_{\odot}$) exhibit anomalous lithium enhancement behavior, which is still an open topic. Given that more massive giants retain more surface lithium, increasing mass by accreting circumstellar matter could be a channel to enrich lithium. We evaluate this process in the current work. Using MESA, we construct a model of matter accretion, including mass loss, that evolves a star from the main sequence turnoff to the red giant branch tip. The mean accretion rate is estimated from the upper limit of the accreted mass and the evolutionary time of the star during this period, and a grid of accretion rates is constructed. We separately consider their effects on the lithium enhancement of giants, both in terms of the mass and the composition of accretion. Accreting matter with higher lithium abundances has a promoting effect on the lithium enhancement of giants. The accreted matter with excess lithium alleviates the dilution of lithium in the convective envelope during the first dredge-up. The added mass results in lower temperatures at the bottom of the convective envelope, which likewise weakens the depletion of surface lithium. Weak accretion of circumstellar matter is a possible route to lithium enhancement for giants, and it predicts an upper limit on the lithium abundance of $\rm \sim 2.5\,dex$. However, the mass increment it requires poses a potential challenge to real astrophysical environments. Such accretion suppresses lithium dilution and depletion of the star during the first dredge-up, thus exhibiting lithium enhancement behavior.

astro-ph.SR↗