语言大模型 / LLM - arXivDaily 专题

2505.21077 2025-10-21 cs.LG cs.AI 90%

Efficient Large Language Model Inference with Neural Block Linearization

Mete Erdogan, Francesco Tonin, Volkan Cevher

机构 * Laboratory for Information and Inference Systems（信息与推断系统实验室）

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

2510.17147 2025-10-21 cs.NI 89%

Mamba4Net: Distilled Hybrid Mamba Large Language Models For Networking

Linhan Xia, Mingzhan Yang, Jingjing Wang, Ziwei Yan, Yakun Ren, Guo Yu, Kai Lei

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

2510.17705 2025-10-21 cs.AI cs.CL 88%

Contextual Attention Modulation: Towards Efficient Multi-Task Adaptation in Large Language Models

Dayan Pan, Zhaoyang Fu, Jingyuan Wang, Xiao Han, Yue Zhu, Xiangyu Zhao

机构 * SCSE, Beihang University（北京航空航天大学系统工程学院）； ERC of ACAT, MOE（教育部人工智能教育联合体）； City University of Hong Kong（香港城市大学）； Huawei Technologies Ltd.（华为技术有限公司）； Key Lab of DIM, MIIT（信息通信技术重点实验室）； SEM, Beihang University（北京航空航天大学电子机械学院）； Zhejiang University of Technology（浙江工业大学）

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

Comments Accepted by CIKM' 25

2506.19505 2025-10-21 cs.CL 88%

AnTKV: Anchor Token-Aware Sub-Bit Vector Quantization for KV Cache in Large Language Models

Zeyu Li, Chuanfu Xiao, Yang Wang, Xiang Liu, Zhenheng Tang, Baotong Lu, Mao Yang, Xinyu Chen, Xiaowen Chu

机构 * DSA Thrust, The Hong Kong University of Science and Technology (Guangzhou)（DSA团队，香港科学与技术大学（广州））； School of Mathematics and Computational Science, Xiangtan University（数学与计算科学学院，湘潭大学）； Microsoft（微软）

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

2501.04762 2025-10-21 cs.IR cs.LG 88%

Efficient and Responsible Adaptation of Large Language Models for Robust and Equitable Top-k Recommendations

Kirandeep Kaur, Vinayak Gupta, Manya Chadha, Chirag Shah

机构 * University of Washington（华盛顿大学）

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

Comments arXiv admin note: text overlap with arXiv:2405.00824

2506.10337 2025-10-21 cs.CV 88%

GeoCAD: Local Geometry-Controllable CAD Generation with Large Language Models

Zhanwei Zhang, Kaiyuan Liu, Junjie Liu, Wenxiao Wang, Binbin Lin, Liang Xie, Chen Shen, Deng Cai

机构 * State Key Lab of CAD&CG（CAD与计算机图形学国家重点实验室）； Alibaba Cloud Computing（阿里云计算）； Zhejiang University of Technology（浙江工业大学）； School of Software Technology, Zhejiang University（浙江大学软件技术学院）

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

Comments Accepted by NeurIPS 2025

2505.19504 2025-10-21 cs.LG cs.AI cs.CL 87%

DOGe: Defensive Output Generation for LLM Protection Against Knowledge Distillation

Pingzhi Li, Zhen Tan, Mohan Zhang, Huaizhi Qu, Huan Liu, Tianlong Chen

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、cs.LG

Comments Code is available at https://github.com/unites-lab/doge

2509.16293 2025-10-21 cs.LG cs.AI cs.DC 86%

Robust LLM Training Infrastructure at ByteDance

Borui Wan, Gaohong Liu, Zuquan Song, Jun Wang, Yun Zhang, Guangming Sheng, Shuguang Wang, Houmin Wei, Chenyuan Wang, Weiqiang Lou, Xi Yang, Mofan Zhang, Kaihua Jiang, Cheng Ren, Xiaoyun Zhi, Menghan Yu, Zhe Nan, Zhuolin Zheng, Baoquan Zhong, Qinlong Wang, Huan Yu, Jinxin Chi, Wang Zhang, Yuhan Li, Zixian Du, Sida Zhao, Yongqiang Zhang, Jingzhe Tang, Zherui Liu, Chuan Wu, Yanghua Peng, Haibin Lin, Wencong Xiao, Xin Liu, Liang Xiang

机构 * The University of Hong Kong（香港大学）； ByteDance Seed（字节跳动种子）

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

2510.16040 2025-10-21 cs.AR cs.AI 85%

Kelle: Co-design KV Caching and eDRAM for Efficient LLM Serving in Edge Computing

Tianhua Xia, Sai Qian Zhang

机构 * Tandon School of Engineering（工程学院）； New York University（纽约大学）

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

2510.17393 2025-10-21 q-fin.PM 85%

3S-Trader: A Multi-LLM Framework for Adaptive Stock Scoring, Strategy, and Selection in Portfolio Optimization

Kefan Chen, Hussain Ahmad, Diksha Goel, Claudia Szabo

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

2510.16418 2025-10-21 cs.DC 85%

FourierCompress: Layer-Aware Spectral Activation Compression for Efficient and Accurate Collaborative LLM Inference

Jian Ma, Xinchen Lyu, Jun Jiang, Longhao Zou, Chenshan Ren, Qimei Cui, Xiaofeng Tao

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

2510.16415 2025-10-21 cs.DC 85%

MeCeFO: Enhancing LLM Training Robustness via Fault-Tolerant Optimization

Rizhen Hu, Yutong He, Ran Yan, Mou Sun, Binghang Yuan, Kun Yuan

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

Comments NeurIPS 2025 poster

2501.06807 2025-10-21 cs.CR 85%

MPCache: MPC-Friendly KV Cache Eviction for Efficient Private LLM Inference

Wenxuan Zeng, Ye Dong, Jinjin Zhou, Jin Tan, Lei Wang, Tao Wei, Runsheng Wang, Meng Li

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

Comments Accepted by NeurIPS 2025

2510.16805 2025-10-21 cs.LG cs.AI 84%

Mixed-Precision Quantization for Language Models: Techniques and Prospects

Mariam Rakka, Marios Fournarakis, Olga Krestinskaya, Jinane Bazzi, Khaled N. Salama, Fadi Kurdahi, Ahmed M. Eltawil, Mohammed E. Fouda

机构 * University of California Irvine（加州大学 Irvine 分校）； Wayve AI ； King Abdullah University of Science and Technology（国王 Abdullah 科学与技术大学）； RAIN AI

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

Comments 46 pages, 6 figures, 5 tables

2510.16290 2025-10-21 cs.CV cs.CL 83%

Cerberus: Real-Time Video Anomaly Detection via Cascaded Vision-Language Models

Yue Zheng, Xiufang Shi, Jiming Chen, Yuanchao Shu

机构 * Zhejiang University of Technology（浙江工业大学）； Zhejiang University（浙江大学）； Hangzhou Dianzi University（杭州电子科技大学）

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

2510.17802 2025-10-21 cs.LG cs.AI math.OC 82%

Unbiased Gradient Low-Rank Projection

Rui Pan, Yang Luo, Yuxing Liu, Yang You, Tong Zhang

机构 * University of Illinois Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）； National University of Singapore（新加坡国立大学）

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

2510.16716 2025-10-21 cs.CR cs.LG 81%

DistilLock: Safeguarding LLMs from Unauthorized Knowledge Distillation on the Edge

Asmita Mohanty, Gezheng Kang, Lei Gao, Murali Annavaram

机构 * University of Southern California（南加州大学）； University of California, Davis（加州大学戴维斯分校）

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

2510.11302 2025-10-21 cs.CV cs.AI cs.LG 81%

When Does Supervised Training Pay Off? The Hidden Economics of Object Detection in the Era of Vision-Language Models

Samer Al-Hamadani

机构 * Automated Manufacturing Department（自动化制造部门）； Al-Khwarizmi College of Engineering（阿尔·卡瓦尔米工程学院）； University of Baghdad（巴格达大学）

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

Comments 30 pages, 12 figures, 4 tables

2510.15945 2025-10-21 cs.LG cs.AI 81%

BEACON: Bayesian Optimal Stopping for Efficient LLM Sampling

Guangya Wan, Zixin Stephen Xu, Sasa Zorc, Manel Baucells, Mengxuan Hu, Hao Wang, Sheng Li

机构 * University of Virginia（弗吉尼亚大学）

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

Comments Under review on ARR

2411.14458 2025-10-21 cs.DC cs.AI cs.LG 81%

Improving training time and GPU utilization in geo-distributed language model training

Palak, Tella Rajashekhar Reddy, Bhaskar Kataria, Rohan Gandhi, Karan Tandon, Debopam Bhattacherjee, Venkata N. Padmanabhan

机构 * Microsoft Research India（微软印度研究院）

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

2510.16968 2025-10-21 cs.LG cs.AI cs.CL 80%

Leave It to the Experts: Detecting Knowledge Distillation via MoE Expert Signatures

Pingzhi Li, Morris Yu-Chao Huang, Zhen Tan, Qingquan Song, Jie Peng, Kai Zou, Yu Cheng, Kaidi Xu, Tianlong Chen

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 Code is at https://github.com/unites-lab/shadow-moe

2510.17197 2025-10-21 cs.CV cs.AI 79%

ZSPAPrune: Zero-Shot Prompt-Aware Token Pruning for Vision-Language Models

Pu Zhang, Yuwei Li, Xingyuan Xian, Guoming Tang

机构 * The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））

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

2510.16292 2025-10-21 cs.LG 79%

QSVD: Efficient Low-rank Approximation for Unified Query-Key-Value Weight Compression in Low-Precision Vision-Language Models

Yutong Wang, Haiyu Wang, Sai Qian Zhang

机构 * Tandon School of Engineering, New York University（纽约大学工程学院）； Courant Institute of Mathematical Sciences, New York University（纽约大学数学科学学院）

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

Comments Accepted as Spotlight paper by NeurIPS 2025

2510.17358 2025-10-21 cs.LG cs.AI 79%

Localist LLMs with Recruitment Learning

Joachim Diederich

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

2510.16718 2025-10-21 cs.SD cs.CL cs.LG 79%

U-Codec: Ultra Low Frame-rate Neural Speech Codec for Fast High-fidelity Speech Generation

Xusheng Yang, Long Zhou, Wenfu Wang, Kai Hu, Shulin Feng, Chenxing Li, Meng Yu, Dong Yu, Yuexian Zou

机构 * Peking University（北京大学）； Tencent AI Lab（腾讯AI实验室）； Tencent Hunyuan（腾讯文生）

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

2510.16552 2025-10-21 cs.LG cs.AI 79%

LANPO: Bootstrapping Language and Numerical Feedback for Reinforcement Learning in LLMs

Ang Li, Yifei Wang, Zhihang Yuan, Stefanie Jegelka, Yisen Wang

机构 * MIT（麻省理工学院）

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

2510.16276 2025-10-21 cs.AI cs.LG 79%

What Limits Agentic Systems Efficiency?

Song Bian, Minghao Yan, Anand Jayarajan, Gennady Pekhimenko, Shivaram Venkataraman

机构 * UW-Madison（威斯康星大学麦迪逊分校）； University of Toronto（多伦多大学）； NVIDIA（英伟达）

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

Comments 27 pages, 15 figures

2510.16045 2025-10-21 cs.LG cs.AI 79%

AMS-QUANT: Adaptive Mantissa Sharing for Floating-point Quantization

Mengtao Lv, Ruiqi Zhu, Xinyu Wang, Yun Li

机构 * Huawei Inc.（华为公司）

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

Comments 12 pages, 6 figures

2510.15992 2025-10-21 cs.LG cs.AI 79%

Stratos: An End-to-End Distillation Pipeline for Customized LLMs under Distributed Cloud Environments

Ziming Dai, Tuo Zhang, Fei Gao, Xingyi Cai, Xiaofei Wang, Cheng Zhang, Wenyu Wang, Chengjie Zang

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

2510.17555 2025-10-21 cs.CL 77%

Language Confusion Gate: Language-Aware Decoding Through Model Self-Distillation

Collin Zhang, Fei Huang, Chenhan Yuan, Junyang Lin

机构 * Alibaba Group（阿里巴巴集团）； Cornell University（康奈尔大学）

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

1. 效率与部署 58 篇

Efficient Large Language Model Inference with Neural Block Linearization

Mamba4Net: Distilled Hybrid Mamba Large Language Models For Networking

Contextual Attention Modulation: Towards Efficient Multi-Task Adaptation in Large Language Models

AnTKV: Anchor Token-Aware Sub-Bit Vector Quantization for KV Cache in Large Language Models

Efficient and Responsible Adaptation of Large Language Models for Robust and Equitable Top-k Recommendations

GeoCAD: Local Geometry-Controllable CAD Generation with Large Language Models

DOGe: Defensive Output Generation for LLM Protection Against Knowledge Distillation

Robust LLM Training Infrastructure at ByteDance

Kelle: Co-design KV Caching and eDRAM for Efficient LLM Serving in Edge Computing

3S-Trader: A Multi-LLM Framework for Adaptive Stock Scoring, Strategy, and Selection in Portfolio Optimization

FourierCompress: Layer-Aware Spectral Activation Compression for Efficient and Accurate Collaborative LLM Inference

MeCeFO: Enhancing LLM Training Robustness via Fault-Tolerant Optimization

MPCache: MPC-Friendly KV Cache Eviction for Efficient Private LLM Inference

Mixed-Precision Quantization for Language Models: Techniques and Prospects

Cerberus: Real-Time Video Anomaly Detection via Cascaded Vision-Language Models

Unbiased Gradient Low-Rank Projection

DistilLock: Safeguarding LLMs from Unauthorized Knowledge Distillation on the Edge

When Does Supervised Training Pay Off? The Hidden Economics of Object Detection in the Era of Vision-Language Models

BEACON: Bayesian Optimal Stopping for Efficient LLM Sampling

Improving training time and GPU utilization in geo-distributed language model training

Leave It to the Experts: Detecting Knowledge Distillation via MoE Expert Signatures

ZSPAPrune: Zero-Shot Prompt-Aware Token Pruning for Vision-Language Models

QSVD: Efficient Low-rank Approximation for Unified Query-Key-Value Weight Compression in Low-Precision Vision-Language Models

Localist LLMs with Recruitment Learning

U-Codec: Ultra Low Frame-rate Neural Speech Codec for Fast High-fidelity Speech Generation

LANPO: Bootstrapping Language and Numerical Feedback for Reinforcement Learning in LLMs

What Limits Agentic Systems Efficiency?

AMS-QUANT: Adaptive Mantissa Sharing for Floating-point Quantization

Stratos: An End-to-End Distillation Pipeline for Customized LLMs under Distributed Cloud Environments

Language Confusion Gate: Language-Aware Decoding Through Model Self-Distillation