语言大模型 / LLM - arXivDaily 专题

2510.19334 2025-10-23 stat.ML cs.AI cs.IR cs.LG 91%

Metadata Extraction Leveraging Large Language Models

Cuize Han, Sesh Jalagam

机构 * Box AI Platform（Box人工智能平台）

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

2408.08878 2025-10-23 cs.HC 90%

Knowledge Prompting: How Knowledge Engineers Use Large Language Models

Elisavet Koutsiana, Johanna Walker, Michelle Nwachukwu, Bohui Zhang, Albert Meroño-Peñuela, Elena Simperl

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

2510.19506 2025-10-23 cs.CL 89%

Lookahead Routing for Large Language Models

Canbin Huang, Tianyuan Shi, Yuhua Zhu, Ruijun Chen, Xiaojun Quan

机构 * School of Computer Science and Engineering, Sun Yat-sen University, China（中山大学计算机科学与工程学院）； Shenzhen Loop Area Institute, China（深圳环湖研究所）

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

2510.19389 2025-10-23 cs.LG 89%

ARA: Adaptive Rank Allocation for Efficient Large Language Model SVD Compression

Lin Xv, Jingsheng Gao, Xian Gao, Ting Liu, Yuzhuo Fu

机构 * Shanghai Jiao Tong University（上海交通大学）

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

2510.19183 2025-10-23 cs.CV cs.AI 88%

PruneHal: Reducing Hallucinations in Multi-modal Large Language Models through Adaptive KV Cache Pruning

Fengyuan Sun, Hui Chen, Xinhao Xu, Dandan Zheng, Jingdong Chen, Jun Zhou, Jungong Han, Guiguang Ding

机构 * School of Software, Tsinghua University（清华大学软件学院）； Ant Group（蚂蚁集团）； Department of Automation, Tsinghua University（清华大学自动化系）

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

2509.20214 2025-10-23 cs.LG cs.AI 88%

Q-Palette: Fractional-Bit Quantizers Toward Optimal Bit Allocation for Efficient LLM Deployment

Deokjae Lee, Hyun Oh Song

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

Comments NeurIPS 2025

2510.08666 2025-10-23 cs.CL cs.AI 84%

dInfer: An Efficient Inference Framework for Diffusion Language Models

Yuxin Ma, Lun Du, Lanning Wei, Kun Chen, Qian Xu, Kangyu Wang, Guofeng Feng, Guoshan Lu, Lin Liu, Xiaojing Qi, Xinyuan Zhang, Zhen Tao, Haibo Feng, Ziyun Jiang, Ying Xu, Zenan Huang, Yihong Zhuang, Haokai Xu, Jiaqi Hu, Zhenzhong Lan, Junbo Zhao, Jianguo Li, Da Zheng

机构 * Ant Group（蚂蚁集团）； Zhejiang University（浙江大学）； Westlake University（西湖大学）； Renmin University of China（中国人民大学）； University of Chinese Academy of Sciences（中国科学院大学）； Shanghai Jiao Tong University（上海交通大学）

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

2510.19307 2025-10-23 cs.CV 82%

Unified Reinforcement and Imitation Learning for Vision-Language Models

Byung-Kwan Lee, Ryo Hachiuma, Yong Man Ro, Yu-Chiang Frank Wang, Yueh-Hua Wu

机构 * NVIDIA ； KAIST（韩国科学技术院）； National Taiwan University（国立台湾大学）

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

Comments NeurIPS 2025, Project page: https://byungkwanlee.github.io/RIL-page

2412.01822 2025-10-23 cs.CV 82%

VLsI: Verbalized Layers-to-Interactions from Large to Small Vision Language Models

Byung-Kwan Lee, Ryo Hachiuma, Yu-Chiang Frank Wang, Yong Man Ro, Yueh-Hua Wu

机构 * NVIDIA ； KAIST（韩国科学技术院）； National Taiwan University（国立台湾大学）

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

Comments CVPR 2025, Project page: https://byungkwanlee.github.io/VLsI-page/

2505.14455 2025-10-23 cs.CL cs.AI 81%

CtrlDiff: Boosting Large Diffusion Language Models with Dynamic Block Prediction and Controllable Generation

Chihan Huang, Hao Tang

机构 * Nanjing University of Science and Technology（南京理工大学）； Peking University（北京大学）

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

2510.18904 2025-10-23 cs.CL cs.AI cs.IR cs.LG 80%

DuoLens: A Framework for Robust Detection of Machine-Generated Multilingual Text and Code

Shriyansh Agrawal, Aidan Lau, Sanyam Shah, Ahan M R, Kevin Zhu, Sunishchal Dev, Vasu Sharma

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

Comments Accepted to 39th Conference on Neural Information Processing Systems (NeurIPS 2025): 4th Workshop on Deep Learning for Code

Journal ref Neural Information Processing Systems (NeurIPS 2025)

2505.08727 2025-10-23 cs.LG cs.AI cs.CL cs.IT math.IT 80%

Memorization-Compression Cycles Improve Generalization

Fangyuan Yu

机构 * Thoughtworks

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

Comments 12 pages, 6 figures, NeurIPS2025 NEGEL Workshop

2510.19801 2025-10-23 cs.LG 79%

The Feasibility of Training Sovereign Language Models in the Global South: A Study of Brazil and Mexico

Sandra Malagon, Monica A. Ulloa Ruiz, Tatiana Elizabeth Sandoval Plaza, Gabriel Rafael Rosario Bolívar, Valentina García Mesa, Ivanna Alvarado Morales

机构 * Carreras con Impacto Aixo（影响深远的课程）

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

Comments 11 pages, 3 figures

2505.18983 2025-10-23 cs.LG cs.CV 79%

AmorLIP: Efficient Language-Image Pretraining via Amortization

Haotian Sun, Yitong Li, Yuchen Zhuang, Niao He, Hanjun Dai, Bo Dai

机构 * Georgia Institute of Technology（佐治亚理工学院）； Swiss Federal Institute of Technology（瑞士联邦理工学院）

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

2510.19498 2025-10-23 cs.LG 77%

Energy-Efficient and Dequantization-Free Q-LLMs: A Spiking Neural Network Approach to Salient Value Mitigation

Chenyu Wang, Zhanglu Yan, Zhi Zhou, Xu Chen, Weng-Fai Wong

机构 * Sun Yat-sen University（中山大学）； National University of Singapore（新加坡国立大学）； Institute for Clarity in Documentation（文档清晰研究所）； Inria Paris-Rocquencourt（巴黎-拉德克利斯研究所）； Rajiv Gandhi University（拉贾·甘地大学）； Tsinghua University（清华大学）； Palmer Research Laboratories（帕默研究实验室）

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

2510.18897 2025-10-23 cs.DC cs.AI cs.DB cs.SE 77%

AI for Distributed Systems Design: Scalable Cloud Optimization Through Repeated LLMs Sampling And Simulators

Jacopo Tagliabue

机构 * Jacopo Tagliabue（独立研究者）

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

Comments Pre-print IAAA workshop submission

2508.12631 2025-10-23 cs.CL 77%

Beyond GPT-5: Making LLMs Cheaper and Better via Performance-Efficiency Optimized Routing

Yiqun Zhang, Hao Li, Jianhao Chen, Hangfan Zhang, Peng Ye, Lei Bai, Shuyue Hu

机构 * Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）

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

Comments This work has been accepted to DAI 2025

2510.19779 2025-10-23 cs.CL cs.AI cs.LG 75%

AdaSPEC: Selective Knowledge Distillation for Efficient Speculative Decoders

Yuezhou Hu, Jiaxin Guo, Xinyu Feng, Tuo Zhao

机构 * University of California, Berkeley（加州大学伯克利分校）； Tsinghua University（清华大学）； Georgia Institute of Technology（佐治亚理工学院）

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

2510.19252 2025-10-23 cs.HC 75%

LLMartini: Seamless and Interactive Leveraging of Multiple LLMs through Comparison and Composition

Yingtian Shi, Jinda Yang, Yuhan Wang, Yiwen Yin, Haoyu Li, Kunyu Gao, Chun Yu

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

2510.18927 2025-10-23 cs.LG cs.AI cs.CL 75%

BAPO: Stabilizing Off-Policy Reinforcement Learning for LLMs via Balanced Policy Optimization with Adaptive Clipping

Zhiheng Xi, Xin Guo, Yang Nan, Enyu Zhou, Junrui Shen, Wenxiang Chen, Jiaqi Liu, Jixuan Huang, Zhihao Zhang, Honglin Guo, Xun Deng, Zhikai Lei, Miao Zheng, Guoteng Wang, Shuo Zhang, Peng Sun, Rui Zheng, Hang Yan, Tao Gui, Qi Zhang, Xuanjing Huang

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

Comments Preprint

2510.18921 2025-10-23 cs.LG cs.AI cs.CL 75%

Benchmarking On-Device Machine Learning on Apple Silicon with MLX

Oluwaseun A. Ajayi, Ogundepo Odunayo

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

Comments 19 pages, 6 figures. Presented at the 6th Deep Learning Indaba (DLI 2024), Dakar, Senegal; non-archival presentation. Poster: https://storage.googleapis.com/indaba-public/Oluwaseun_Ajayi%20.pdf

2510.18527 2025-10-23 cs.IR 75%

LLMs as Sparse Retrievers:A Framework for First-Stage Product Search

Hongru Song, Yu-an Liu, Ruqing Zhang, Jiafeng Guo, Maarten de Rijke, Sen Li, Wenjun Peng, Fuyu Lv, Xueqi Cheng

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

Comments 16 pages

2510.19327 2025-10-23 cs.MA cs.AI 74%

SORA-ATMAS: Adaptive Trust Management and Multi-LLM Aligned Governance for Future Smart Cities

Usama Antuley, Shahbaz Siddiqui, Sufian Hameed, Waqas Arif, Subhan Shah, Syed Attique Shah

机构 * organization= Department of Computer Science, National University of Computer \& Emerging Sciences , addressline= St-4 Sector 17-D On National Highway , city= Karachi , postcode= 75160 , state= , country= Pakistan ； organization= Balochistan University of Information Technology, Engineering ； organization= Department of Computer Science, Birmingham City University , addressline= STEAMhouse, Belmont Row , city= Birmingham , postcode= B4 7RQ , country= United Kingdom

专题命中效率与部署：LLM(title)；分类 cs.AI

2510.19796 2025-10-23 cs.LG cs.CL 73%

Blackbox Model Provenance via Palimpsestic Membership Inference

Rohith Kuditipudi, Jing Huang, Sally Zhu, Diyi Yang, Christopher Potts, Percy Liang

机构 * Stanford University（斯坦福大学）

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

2510.18279 2025-10-23 cs.CL cs.AI 73%

Text or Pixels? It Takes Half: On the Token Efficiency of Visual Text Inputs in Multimodal LLMs

Yanhong Li, Zixuan Lan, Jiawei Zhou

机构 * Allen Institute for AI（艾伦人工智能研究所）； University of Chicago（芝加哥大学）； Stony Brook University（石溪大学）

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

Comments Accepted to EMNLP 2025 Findings ("Text or Pixels? Evaluating Efficiency and Understanding of LLMs with Visual Text Inputs")

2503.07663 2025-10-23 cs.LG cs.AI 73%

Merge then Realign: Simple and Effective Modality-Incremental Continual Learning for Multimodal LLMs

Dingkun Zhang, Shuhan Qi, Xinyu Xiao, Kehai Chen, Xuan Wang

机构 * Harbin Institute of Technology, Shenzhen（哈尔滨工业大学（深圳））； Guangdong Provincial Key Laboratory of Novel Security Intelligence Technologies（广东省新型安全智能技术重点实验室）

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

Comments EMNLP 2025 Main Conference

2510.17519 2025-10-23 cs.CV cs.AI 70%

MUG-V 10B: High-efficiency Training Pipeline for Large Video Generation Models

Yongshun Zhang, Zhongyi Fan, Yonghang Zhang, Zhangzikang Li, Weifeng Chen, Zhongwei Feng, Chaoyue Wang, Peng Hou, Anxiang Zeng

机构 * LLM Team, Shopee Pte. Ltd.（Shopee 股份有限公司语言模型团队）

专题命中效率与部署：pretraining(abstract)；post-training(abstract)；分类 cs.AI

Comments Technical Report; Project Page: https://github.com/Shopee-MUG/MUG-V

2506.13579 2025-10-23 cs.LG cs.AI cs.CL cs.CV 67%

Flexible-length Text Infilling for Discrete Diffusion Models

Andrew Zhang, Anushka Sivakumar, Chiawei Tang, Chris Thomas

机构 * Department of Computer Science（计算机科学系）

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

Comments Major edit of methodology section. Matches EMNLP camera-ready version

2508.08514 2025-10-23 cs.CL cs.LG 62%

DeCAL Tokenwise Compression

Sameer Panwar

机构 * Google DeepMind（谷歌DeepMind）

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

2510.19773 2025-10-23 cs.LG cs.CR 57%

The Tail Tells All: Estimating Model-Level Membership Inference Vulnerability Without Reference Models

Euodia Dodd, Nataša Krčo, Igor Shilov, Yves-Alexandre de Montjoye

机构 * Imperial College London（伦敦帝国学院）

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

1. 效率与部署 36 篇

Metadata Extraction Leveraging Large Language Models

Knowledge Prompting: How Knowledge Engineers Use Large Language Models

Lookahead Routing for Large Language Models

ARA: Adaptive Rank Allocation for Efficient Large Language Model SVD Compression

PruneHal: Reducing Hallucinations in Multi-modal Large Language Models through Adaptive KV Cache Pruning

Q-Palette: Fractional-Bit Quantizers Toward Optimal Bit Allocation for Efficient LLM Deployment

dInfer: An Efficient Inference Framework for Diffusion Language Models

Unified Reinforcement and Imitation Learning for Vision-Language Models

VLsI: Verbalized Layers-to-Interactions from Large to Small Vision Language Models

CtrlDiff: Boosting Large Diffusion Language Models with Dynamic Block Prediction and Controllable Generation

DuoLens: A Framework for Robust Detection of Machine-Generated Multilingual Text and Code

Memorization-Compression Cycles Improve Generalization

The Feasibility of Training Sovereign Language Models in the Global South: A Study of Brazil and Mexico

AmorLIP: Efficient Language-Image Pretraining via Amortization

Energy-Efficient and Dequantization-Free Q-LLMs: A Spiking Neural Network Approach to Salient Value Mitigation

AI for Distributed Systems Design: Scalable Cloud Optimization Through Repeated LLMs Sampling And Simulators

Beyond GPT-5: Making LLMs Cheaper and Better via Performance-Efficiency Optimized Routing

AdaSPEC: Selective Knowledge Distillation for Efficient Speculative Decoders

LLMartini: Seamless and Interactive Leveraging of Multiple LLMs through Comparison and Composition

BAPO: Stabilizing Off-Policy Reinforcement Learning for LLMs via Balanced Policy Optimization with Adaptive Clipping

Benchmarking On-Device Machine Learning on Apple Silicon with MLX

LLMs as Sparse Retrievers:A Framework for First-Stage Product Search

SORA-ATMAS: Adaptive Trust Management and Multi-LLM Aligned Governance for Future Smart Cities

Blackbox Model Provenance via Palimpsestic Membership Inference

Text or Pixels? It Takes Half: On the Token Efficiency of Visual Text Inputs in Multimodal LLMs

Merge then Realign: Simple and Effective Modality-Incremental Continual Learning for Multimodal LLMs

MUG-V 10B: High-efficiency Training Pipeline for Large Video Generation Models

Flexible-length Text Infilling for Discrete Diffusion Models

DeCAL Tokenwise Compression

The Tail Tells All: Estimating Model-Level Membership Inference Vulnerability Without Reference Models