语言大模型 / LLM - arXivDaily 专题

2505.21334 2025-10-13 cs.CV 89%

HoliTom: Holistic Token Merging for Fast Video Large Language Models

Kele Shao, Keda Tao, Can Qin, Haoxuan You, Yang Sui, Huan Wang

机构 * Zhejiang University（浙江大学）； Shanghai Innovation Institute（上海创新研究院）； Westlake University（西湖大学）； Salesforce AI Research（Salesforce AI研究院）； Columbia University（哥伦比亚大学）； Rice University（Rice大学）

专题命中效率与部署：large language model(title,abstract)；language model(title,abstract)；LLM(abstract)

Comments code link: https://github.com/cokeshao/HoliTom

2510.09103 2025-10-13 cs.LG 87%

AdaPM: a Partial Momentum Algorithm for LLM Training

Yimu Zhang, Yuanshi Liu, Cong Fang

机构 * State Key Lab of General AI, School of Intelligence Science and Technology, Peking University（通用人工智能国家重点实验室，智能科学与技术学院，北京大学）； Institute for Artificial Intelligence, Peking University（人工智能研究院，北京大学）

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)；pretraining(abstract)

2410.12444 2025-10-13 cs.CL 86%

Augmenting Compliance-Guaranteed Customer Service Chatbots: Context-Aware Knowledge Expansion with Large Language Models

Mengze Hong, Chen Jason Zhang, Di Jiang, Yuanqin He

机构 * Hong Kong Polytechnic University（香港理工大学）； AI Group, WeBank Co., Ltd（WeBank金融科技部）

专题命中效率与部署：large language model(title)；language model(title)；LLM(abstract)；分类 cs.CL

Comments Accepted by EMNLP 2025 Industry Track

2510.09332 2025-10-13 cs.CL cs.AI 86%

FLRC: Fine-grained Low-Rank Compressor for Efficient LLM Inference

Yu-Chen Lu, Chong-Yan Chen, Chi-Chih Chang, Yu-Fang Hu, Kai-Chiang Wu

机构 * National Yang Ming Chiao Tung University ； Macronix International Co., Ltd. ； Cornell University

专题命中效率与部署：LLM(title,abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL、cs.AI

Comments Accepted by EMNLP 2025

2510.08758 2025-10-13 stat.ME cs.CL cs.LG stat.AP 86%

A Design-based Solution for Causal Inference with Text: Can a Language Model Be Too Large?

Graham Tierney, Srikar Katta, Christopher Bail, Sunshine Hillygus, Alexander Volfovsky

机构 * Duke University（杜克大学）； Netflix

专题命中效率与部署：language model(title,abstract)；LLM(abstract)；large language model(abstract)；分类 cs.CL、cs.LG

2510.08623 2025-10-13 cs.CL cs.LG 86%

PARSE: LLM Driven Schema Optimization for Reliable Entity Extraction

Anubhav Shrimal, Aryan Jain, Soumyajit Chowdhury, Promod Yenigalla

机构 * RBS Tech Sciences, Amazon（RBS科技科学，亚马逊）

专题命中效率与部署：LLM(title,abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL、cs.LG

Comments EMNLP 2025 Industry Track

2509.24840 2025-10-13 cs.LG cs.CE 85%

Cell2Text: Multimodal LLM for Generating Single-Cell Descriptions from RNA-Seq Data

Oussama Kharouiche, Aris Markogiannakis, Xiao Fei, Michail Chatzianastasis, Michalis Vazirgiannis

机构 * École Polytechnique, IP Paris（巴黎理工学院，IP巴黎）

专题命中效率与部署：LLM(title)；large language model(abstract)；language model(abstract)；foundation model(abstract)

2510.08600 2025-10-13 cs.CL cs.AI cs.LG 85%

Recover-LoRA: Data-Free Accuracy Recovery of Degraded Language Models via Low-Rank Adaptation

Devleena Das, Rajeev Patwari, Ashish Sirasao

机构 * Advanced Micro Devices, Inc. (AMD)（先进微器件公司（AMD））

专题命中效率与部署：language model(title,abstract)；small language model(abstract)；分类 cs.CL、cs.AI、cs.LG

Comments Accepted to EMNLP 2025 Industry Track

2510.09004 2025-10-13 cs.CL 84%

Decoupling Safety into Orthogonal Subspace: Cost-Efficient and Performance-Preserving Alignment for Large Language Models

Yutao Mou, Xiaoling Zhou, Yuxiao Luo, Shikun Zhang, Wei Ye

机构 * National Engineering Research Center for Software Engineering, Peking University, China（软件工程国家工程研究中心，北京大学，中国）

专题命中效率与部署：large language model(title)；language model(title)；分类 cs.CL

Comments Work in Progress

2510.08610 2025-10-13 cs.SE cs.AI cs.LG 84%

Relative Positioning Based Code Chunking Method For Rich Context Retrieval In Repository Level Code Completion Task With Code Language Model

Imranur Rahman, Md Rayhanur Rahman

机构 * North Carolina State University（北卡罗来纳州立大学）； The University of Alabama（阿拉巴马大学）

专题命中效率与部署：language model(title,abstract)；large language model(abstract)；分类 cs.AI、cs.LG

Comments Accepted to Context Collection Workshop co-located with ASE 2025

2412.04787 2025-10-13 cs.LG cs.CL 84%

Direct Quantized Training of Language Models with Stochastic Rounding

Kaiyan Zhao, Tsuguchika Tabaru, Kenichi Kobayashi, Takumi Honda, Masafumi Yamazaki, Yoshimasa Tsuruoka

机构 * The University of Tokyo（东京大学）； Fujitsu Limited（富士通有限公司）

专题命中效率与部署：language model(title,abstract)；large language model(abstract)；分类 cs.CL、cs.LG

Comments Accepted to ACML 2025

2510.00071 2025-10-13 cs.AI cs.CL 82%

ARS: Adaptive Reasoning Suppression for Efficient Large Reasoning Language Models

Dongqi Zheng

机构 * Independent Researcher（独立研究者）

专题命中效率与部署：language model(title,abstract)；分类 cs.CL、cs.AI

Comments Accepted by 39th NeurIPS - Foundations of Reasoning in Language Models

2502.15260 2025-10-13 cs.CL 81%

LightMamba: Efficient Mamba Acceleration on FPGA with Quantization and Hardware Co-design

Renjie Wei, Songqiang Xu, Linfeng Zhong, Zebin Yang, Qingyu Guo, Yuan Wang, Runsheng Wang, Meng Li

机构 * Institute for Artificial Intelligence（人工智能研究院）； School of Integrated Circuits（集成电路学院）； Beijing Advanced Innovation Center for Integrated Circuits（北京集成电路先进创新中心）； Institute of Electronic Design Automation（电子设计自动化研究院）； School of Software and Microelectronics（软件与微电子学院）； School of Electronic and Computer Engineering（电子与计算机工程学院）

专题命中效率与部署：LLM(abstract)；large language model(abstract)；language model(abstract)；post-training(abstract)

Comments Accepted by DATE 2025

2510.03298 2025-10-13 cs.LG cs.CL cs.DC 81%

CAFL-L: Constraint-Aware Federated Learning with Lagrangian Dual Optimization for On-Device Language Models

Dongqi Zheng, Wenjin Fu

机构 * Purdue University（普渡大学）； Carnegie Mellon University（卡内基梅隆大学）

专题命中效率与部署：language model(title,abstract)；分类 cs.CL、cs.LG

Comments Accepted by 39th NeurIPS - Constrained Optimization for Machine Learning

2510.08659 2025-10-13 cs.LG cs.AI 81%

Provably Robust Adaptation for Language-Empowered Foundation Models

Yuni Lai, Xiaoyu Xue, Linghui Shen, Yulun Wu, Gaolei Li, Song Guo, Kai Zhou, Bin Xiao

机构 * Department of Computing, The Hong Kong Polytechnic University（香港理工大学计算机系）； College of Systems and Engineering, National University of Defense Technology（国防科技大学系统工程学院）； School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University（上海交通大学电子信息与电气工程学院）； Department of Computer Science and Engineering, Hong Kong University of Science and Technology（香港科技大学计算机科学与工程系）

专题命中效率与部署：foundation model(title,abstract)；分类 cs.AI、cs.LG

Comments 19 pages

2509.06836 2025-10-13 cs.CL cs.AI cs.LG 80%

COMPACT: Common-token Optimized Model Pruning Across Channels and Tokens

Eugene Kwek, Wenpeng Yin

机构 * Department of Computer Science & Engineering（计算机科学与工程系）

专题命中效率与部署：large language model(abstract)；language model(abstract)；small language model(abstract)；分类 cs.CL、cs.AI、cs.LG

2510.08774 2025-10-13 cs.LG cs.AI cs.CL 80%

Struc-EMB: The Potential of Structure-Aware Encoding in Language Embeddings

Shikun Liu, Haoyu Wang, Mufei Li, Pan Li

机构 * Georgia Institute of Technology（佐治亚理工学院）

专题命中效率与部署：LLM(abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL、cs.AI、cs.LG

2510.08601 2025-10-13 cs.CL cs.AI cs.LG cs.MA 80%

Mnemosyne: An Unsupervised, Human-Inspired Long-Term Memory Architecture for Edge-Based LLMs

Aneesh Jonelagadda, Christina Hahn, Haoze Zheng, Salvatore Penachio

机构 * Kaliber AI, Research Division（Kaliber AI 研究部门）

专题命中效率与部署：LLM(abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL、cs.AI、cs.LG

Comments 12 pages, 4 figures

2510.01795 2025-10-13 cs.RO cs.AI 79%

Nav-EE: Navigation-Guided Early Exiting for Efficient Vision-Language Models in Autonomous Driving

Haibo Hu, Lianming Huang, Xinyu Wang, Yufei Cui, Shangyu Wu, Nan Guan, Chun Jason Xue

机构 * Department of Computer Science, City University of Hong Kong（香港城市大学计算机科学系）； Department of Computer Science, McGill University（麦吉尔大学计算机科学系）； Department of Computer Science, Mohamed bin Zayed University of Artificial Intelligence（马尔代夫穆罕默德· bin·扎耶德人工智能大学计算机科学系）

专题命中效率与部署：language model(title,abstract)；分类 cs.AI

2506.14087 2025-10-13 cs.LG 79%

Multi-Scale Finetuning for Encoder-based Time Series Foundation Models

Zhongzheng Qiao, Chenghao Liu, Yiming Zhang, Ming Jin, Quang Pham, Qingsong Wen, P. N. Suganthan, Xudong Jiang, Savitha Ramasamy

机构 * Nanyang Technological University（南洋理工大学）； Institute for Infocomm Research, A*STAR（信息通信研究所，A*STAR）； Salesforce AI Research（Salesforce人工智能研究）； Griffith University（格里菲斯大学）； Squirrel Ai Learning（Squirrel AI学习）； Qatar University（卡塔尔大学）

专题命中效率与部署：foundation model(title,abstract)；分类 cs.LG

Comments Accepted by NeurIPS 2025

2506.05404 2025-10-13 cs.CV cs.AI 79%

AD-EE: Early Exiting for Fast and Reliable Vision-Language Models in Autonomous Driving

Lianming Huang, Haibo Hu, Yufei Cui, Jiacheng Zuo, Shangyu Wu, Nan Guan, Chun Jason Xue

专题命中效率与部署：language model(title,abstract)；分类 cs.AI

Comments We believe that the contribution of this paper is not enough, so we integrated it into another new paper. The arXiv ID of the new paper is arXiv:2510.01795

2510.02463 2025-10-13 cs.CL cs.AI cs.MA 79%

CLARITY: Clinical Assistant for Routing, Inference, and Triage

Vladimir Shaposhnikov, Aleksandr Nesterov, Ilia Kopanichuk, Ivan Bakulin, Egor Zhelvakov, Ruslan Abramov, Ekaterina Tsapieva, Iaroslav Bespalov, Dmitry V. Dylov, Ivan Oseledets

机构 * AIRI ； Skoltech（斯克里普奇克技术学院）； MIPT（莫斯科理工学院）； SberMedAI ； Sber

专题命中效率与部署：LLM(abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL、cs.AI

Comments Accepted to EMNLP 2025 (Industrial Track)

2510.08731 2025-10-13 cs.ET cs.AI cs.CL cs.SY eess.SY 79%

When to Reason: Semantic Router for vLLM

Chen Wang, Xunzhuo Liu, Yuhan Liu, Yue Zhu, Xiangxi Mo, Junchen Jiang, Huamin Chen

机构 * IBM Research（IBM研究院）； Tencent（腾讯）； University of Chicago（芝加哥大学）； UC Berkeley（伯克利大学）； Red Hat（红帽公司）

专题命中效率与部署：LLM(abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL、cs.AI

Comments 5 pages, excluding references and appendix. To be appeared at Workshop on ML for Systems at NeurIPS 2025, December 6, 2025 https://mlforsystems.org/

2510.08813 2025-10-13 cs.CL cs.CR 77%

The Model's Language Matters: A Comparative Privacy Analysis of LLMs

Abhishek K. Mishra, Antoine Boutet, Lucas Magnana

机构 * Inria（法国国家信息与自动化研究所）； INSA Lyon（里昂国立应用科学学院）； CITI（信息与技术研究所）

专题命中效率与部署：LLM(abstract)；large language model(abstract)；language model(abstract)；分类 cs.CL

2510.08589 2025-10-13 cs.CV cs.AI 77%

Beyond CNNs: Efficient Fine-Tuning of Multi-Modal LLMs for Object Detection on Low-Data Regimes

Nirmal Elamon, Rouzbeh Davoudi

机构 * Artificial Creative intelligence (ACI)（人工创意智能（ACI））； Expedia Group（Expedia集团）

专题命中效率与部署：LLM(abstract)；large language model(abstract)；language model(abstract)；分类 cs.AI

2510.08605 2025-10-13 cs.CL cs.AI cs.CR cs.LG 75%

Toward a Safer Web: Multilingual Multi-Agent LLMs for Mitigating Adversarial Misinformation Attacks

Nouar Aldahoul, Yasir Zaki

机构 * New York University Abu Dhabi（纽约大学阿布扎比分校）

专题命中效率与部署：large language model(abstract)；language model(abstract)；分类 cs.CL、cs.AI、cs.LG

2510.08798 2025-10-13 cs.CL 74%

Learning What to Remember: Adaptive Probabilistic Memory Retention for Memory-Efficient Language Models

S M Rafiuddin, Muntaha Nujat Khan

专题命中效率与部署：language model(title)；分类 cs.CL

Comments 14 Pages, 2 Figures, 6 Table, Accepted at EMNLP 2025 Findings as a Short Paper

2505.20318 2025-10-13 cs.CL cs.AI 73%

Beyond Demonstrations: Dynamic Vector Construction from Latent Representations

Wang Cai, Hsiu-Yuan Huang, Zhixiang Wang, Yunfang Wu

机构 * National Key Laboratory for Multimedia Information Processing, Peking University（国家多媒体信息处理重点实验室，北京大学）； MOE Key Laboratory of Computational Linguistics, Peking University（教育部计算语言学重点实验室，北京大学）； School of Software and Microelectronics, Peking University（软件与微电子学院，北京大学）； School of Computer Science, Peking University（计算机科学学院，北京大学）

专题命中效率与部署：large language model(abstract)；language model(abstract)；分类 cs.CL、cs.AI

Comments 9 pages, 4 figures. Accepted to EMNLP 2025 (Main Conference)

2502.01156 2025-10-13 cs.IT math.IT 71%

On the impact of the parametrization of deep convolutional neural networks on post-training quantization

Samy Houache, Jean François Aujol, Yann Traonmilin

专题命中效率与部署：post-training(title)

2510.09309 2025-10-13 cs.CL 70%

Mask Tokens as Prophet: Fine-Grained Cache Eviction for Efficient dLLM Inference

Jianuo Huang, Yaojie Zhang, Yicun Yang, Benhao Huang, Biqing Qi, Dongrui Liu, Linfeng Zhang

机构 * School of Artificial Intelligence, Shanghai Jiao Tong University（上海交通大学人工智能学院）； Huazhong University of Science and Technology（华中科技大学）； University of Electronic Science and Technology of China（电子科技大学）； Carnegie Mellon University（卡内基梅隆大学）； Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）

专题命中效率与部署：large language model(abstract)；language model(abstract)；分类 cs.CL

Comments 17 pages, 8 figures

1. 效率与部署 44 篇

HoliTom: Holistic Token Merging for Fast Video Large Language Models

AdaPM: a Partial Momentum Algorithm for LLM Training

Augmenting Compliance-Guaranteed Customer Service Chatbots: Context-Aware Knowledge Expansion with Large Language Models

FLRC: Fine-grained Low-Rank Compressor for Efficient LLM Inference

A Design-based Solution for Causal Inference with Text: Can a Language Model Be Too Large?

PARSE: LLM Driven Schema Optimization for Reliable Entity Extraction

Cell2Text: Multimodal LLM for Generating Single-Cell Descriptions from RNA-Seq Data

Recover-LoRA: Data-Free Accuracy Recovery of Degraded Language Models via Low-Rank Adaptation

Decoupling Safety into Orthogonal Subspace: Cost-Efficient and Performance-Preserving Alignment for Large Language Models

Relative Positioning Based Code Chunking Method For Rich Context Retrieval In Repository Level Code Completion Task With Code Language Model

Direct Quantized Training of Language Models with Stochastic Rounding

ARS: Adaptive Reasoning Suppression for Efficient Large Reasoning Language Models

LightMamba: Efficient Mamba Acceleration on FPGA with Quantization and Hardware Co-design

CAFL-L: Constraint-Aware Federated Learning with Lagrangian Dual Optimization for On-Device Language Models

Provably Robust Adaptation for Language-Empowered Foundation Models

COMPACT: Common-token Optimized Model Pruning Across Channels and Tokens

Struc-EMB: The Potential of Structure-Aware Encoding in Language Embeddings

Mnemosyne: An Unsupervised, Human-Inspired Long-Term Memory Architecture for Edge-Based LLMs

Nav-EE: Navigation-Guided Early Exiting for Efficient Vision-Language Models in Autonomous Driving

Multi-Scale Finetuning for Encoder-based Time Series Foundation Models

AD-EE: Early Exiting for Fast and Reliable Vision-Language Models in Autonomous Driving

CLARITY: Clinical Assistant for Routing, Inference, and Triage

When to Reason: Semantic Router for vLLM

The Model's Language Matters: A Comparative Privacy Analysis of LLMs

Beyond CNNs: Efficient Fine-Tuning of Multi-Modal LLMs for Object Detection on Low-Data Regimes

Toward a Safer Web: Multilingual Multi-Agent LLMs for Mitigating Adversarial Misinformation Attacks

Learning What to Remember: Adaptive Probabilistic Memory Retention for Memory-Efficient Language Models

Beyond Demonstrations: Dynamic Vector Construction from Latent Representations

On the impact of the parametrization of deep convolutional neural networks on post-training quantization

Mask Tokens as Prophet: Fine-Grained Cache Eviction for Efficient dLLM Inference