语言大模型 / LLM - arXivDaily 专题

2509.05320 2025-09-09 cs.CR cs.LG 89%

Privacy-Preserving Offloading for Large Language Models in 6G Vehicular Networks

Ikhlasse Badidi, Nouhaila El Khiyaoui, Aya Riany, Badr Ben Elallid, Amine Abouaomar

机构 * School of Science and Engineering, Al Akhawayn University in Ifrane（科学与工程学院，阿尔阿赫韦因大学）； Department of Electrical and Computer Engineering, Université du Québec à Trois-Rivières（电气与计算机工程系，魁北克大学Trois-Rivières分校）

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

Comments 7 pages, 6 figures, 1 algorithm, 5 equations

2406.07831 2025-09-09 cs.LG 88%

ALPS: Improved Optimization for Highly Sparse One-Shot Pruning for Large Language Models

Xiang Meng, Kayhan Behdin, Haoyue Wang, Rahul Mazumder

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

2509.05852 2025-09-09 stat.ML cs.LG math.ST stat.TH 88%

Fisher Random Walk: Automatic Debiasing Contextual Preference Inference for Large Language Model Evaluation

Yichi Zhang, Alexander Belloni, Ethan X. Fang, Junwei Lu, Xiaoan Xu

机构 * Department of Statistics, Indiana University Bloomington（印第安纳大学布卢明顿分校统计学系）； Fuqua School of Business, Duke University（杜克大学福克商学院）； Department of Biostatistics & Bioinformatics, Duke University（杜克大学生物统计学与生物信息学系）； Department of Biostatistics, Harvard T.H. Chan School of Public Health（哈佛大学T.H. Chan公共卫生学院生物统计学系）

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

2509.01190 2025-09-09 cs.CL 88%

Efficient Large Language Models with Zero-Shot Adjustable Acceleration

Sajjad Kachuee, Mohammad Sharifkhani

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

2509.05936 2025-09-09 cs.NI cs.LG 85%

ALPHA: LLM-Enabled Active Learning for Human-Free Network Anomaly Detection

Xuanhao Luo, Shivesh Madan Nath Jha, Akruti Sinha, Zhizhen Li, Yuchen Liu

机构 * North Carolina State University, USA（北卡罗来纳州立大学）

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

Comments Accepted at 44th IEEE International Performance Computing and Communications Conference (IPCCC 2025)

2506.15006 2025-09-09 cs.AR cs.AI cs.DC cs.ET cs.PF 85%

Scaling Intelligence: Designing Data Centers for Next-Gen Language Models

Jesmin Jahan Tithi, Hanjiang Wu, Avishaii Abuhatzera, Fabrizio Petrini

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

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

Comments 14 pages, submitted to SC25 for review

2509.05263 2025-09-09 cs.AI cs.CV cs.LG 85%

LatticeWorld: A Multimodal Large Language Model-Empowered Framework for Interactive Complex World Generation

Yinglin Duan, Zhengxia Zou, Tongwei Gu, Wei Jia, Zhan Zhao, Luyi Xu, Xinzhu Liu, Yenan Lin, Hao Jiang, Kang Chen, Shuang Qiu

机构 * NetEase, Inc.（网易公司）； Beihang University（北京航空航天大学）； Tsinghua University（清华大学）； City University of Hong Kong（香港城市大学）； Independent Researcher & Technical Artists（独立研究者及技术艺术家）

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

2509.05584 2025-09-09 cs.LG cs.CV cs.PF 84%

ProfilingAgent: Profiling-Guided Agentic Reasoning for Adaptive Model Optimization

Sadegh Jafari, Aishwarya Sarkar, Mohiuddin Bilwal, Ali Jannesari

机构 * Department of Computer Science（计算机科学系）； Iowa State University（爱荷华州立大学）

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

Comments 13 pages, 3 figures, 5 tables, 1 algorithm

2509.06768 2025-09-09 cs.RO 82%

Embodied Hazard Mitigation using Vision-Language Models for Autonomous Mobile Robots

Oluwadamilola Sotomi, Devika Kodi, Kiruthiga Chandra Shekar, Aliasghar Arab

机构 * Department of Mechanical and Aerospace Engineering, Tandon School of Engineering, New York University（机械与航空航天工程系，坦顿工程学院，纽约大学）； GenAuto.ai by General Autonomy Inc.（General Autonomy Inc. 的 GenAuto.ai）

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

2509.06321 2025-09-09 cs.CV 82%

Text4Seg++: Advancing Image Segmentation via Generative Language Modeling

Mengcheng Lan, Chaofeng Chen, Jiaxing Xu, Zongrui Li, Yiping Ke, Xudong Jiang, Yingchen Yu, Yunqing Zhao, Song Bai

机构 * College of Computing and Data Science, Nanyang Technological University（computing and Data Science学院，南洋理工大学）； School of Electrical and Electronic Engineering, Nanyang Technological University（Electrical and Electronic Engineering学院，南洋理工大学）； School of Artificial Intelligence, Wuhan University（Artificial Intelligence学院，武汉大学）； ByteDance（字节跳动）

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

Comments Extended version of our conference paper arXiv:2410.09855

2509.05962 2025-09-09 cs.HC 82%

The Reel Deal: Designing and Evaluating LLM-Generated Short-Form Educational Videos

Lazaros Stavrinou, Argyris Constantinides, Marios Belk, Vasos Vassiliou, Fotis Liarokapis, Marios Constantinides

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

Comments 9 pages, 3 figures, 1 table

2509.05925 2025-09-09 cs.CV cs.IT math.IT 82%

Compression Beyond Pixels: Semantic Compression with Multimodal Foundation Models

Ruiqi Shen, Haotian Wu, Wenjing Zhang, Jiangjing Hu, Deniz Gunduz

机构 * Department of Electrical and Electronic Engineering, Imperial College London（帝国理工学院电子与电气工程系）

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

Comments Published as a conference paper at IEEE 35th Workshop on Machine Learning for Signal Processing (MLSP)

2509.05355 2025-09-09 cs.RO cs.MA 82%

Human-LLM Synergy in Context-Aware Adaptive Architecture for Scalable Drone Swarm Operation

Ahmed R. Sadik, Muhammad Ashfaq, Niko Mäkitalo, Tommi Mikkonen

机构 * Honda Research Institute Europe, Germany（本田欧洲研究机构）

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

2509.06631 2025-09-09 cs.CL 81%

Guided Decoding and Its Critical Role in Retrieval-Augmented Generation

Özgür Uğur, Musa Yılmaz, Esra Şavirdi, Özay Ezerceli, Mahmut El Huseyni, Selva Taş, Reyhan Bayraktar

机构 * Newmind AI Istanbul, Türkiye（新 mind AI 伊斯坦布尔，土耳其）

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

2509.06518 2025-09-09 cs.CL cs.AI 81%

Crown, Frame, Reverse: Layer-Wise Scaling Variants for LLM Pre-Training

Andrei Baroian, Kasper Notebomer

机构 * LIACS, Leiden University, The Netherlands（LIACS，莱顿大学，荷兰）

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

Comments The reported results are skewed due to a data type mismatch. The dataset was saved with int32, but the data loader interpreted it as uint16. As a result, each 32-bit token was incorrectly split into two 16-bit tokens. Outcome: a consistent artifact where every other token is zero

2509.05333 2025-09-09 cs.CV cs.AI 79%

RT-VLM: Re-Thinking Vision Language Model with 4-Clues for Real-World Object Recognition Robustness

Junghyun Park, Tuan Anh Nguyen, Dugki Min

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

2506.08727 2025-09-09 cs.LG cs.AI cs.CY cs.SE 79%

Breaking the ICE: Exploring promises and challenges of benchmarks for Inference Carbon & Energy estimation for LLMs

Samarth Sikand, Rohit Mehra, Priyavanshi Pathania, Nikhil Bamby, Vibhu Saujanya Sharma, Vikrant Kaulgud, Sanjay Podder, Adam P. Burden

机构 * Accenture Labs, India（Accenture实验室，印度）； Accenture, India（Accenture，印度）； Accenture, USA（Accenture，美国）

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 5 pages. To be published in the proceedings of 9th International Workshop on Green and Sustainable Software (GREENS '25), April 29, 2025, Ottawa, Canada (Co-located with ICSE 2025)

2509.05576 2025-09-09 cs.CV 78%

Sensitivity-Aware Post-Training Quantization for Deep Neural Networks

Zekang Zheng, Haokun Li, Yaofo Chen, Mingkui Tan, Qing Du

机构 * South China University of Technology（华南理工大学）； Pazhou Laboratory（Pazhou实验室）； Key Laboratory of Big Data and Intelligent Robot, Ministry of Education（大数据与智能机器人重点实验室，教育部）

专题命中效率与部署：post-training(title,abstract)

Comments Accepted by PRCV 2025

2509.06338 2025-09-09 cs.CR cs.LG 77%

Embedding Poisoning: Bypassing Safety Alignment via Embedding Semantic Shift

Shuai Yuan, Zhibo Zhang, Yuxi Li, Guangdong Bai, Wang Kailong

机构 * University of Electronic Science and Technology of China（电子科技大学）； Huazhong University of Science and Technology（华中科技大学）； The University of Queensland（昆士兰大学）

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

Comments 16 pages,9 figures

2504.03165 2025-09-09 cs.CL 77%

Efficient Dynamic Clustering-Based Document Compression for Retrieval-Augmented-Generation

Weitao Li, Kaiming Liu, Xiangyu Zhang, Xuanyu Lei, Weizhi Ma, Yang Liu

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

2509.05883 2025-09-09 cs.CR cs.AI 77%

Multimodal Prompt Injection Attacks: Risks and Defenses for Modern LLMs

Andrew Yeo, Daeseon Choi

机构 * Ranchview High School（拉文斯维尔高中）； Soongsil University（松山大学）

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

Comments 8 pages, 4 figures, 2 tables

2509.06362 2025-09-09 cs.DC 75%

MaaSO: SLO-aware Orchestration of Heterogeneous Model Instances for MaaS

Mo Xuan, Zhang yue, Wu Weigang

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

2509.06069 2025-09-09 econ.GN cs.HC q-fin.EC 75%

From Digital Distrust to Codified Honesty: Experimental Evidence on Generative AI in Credence Goods Markets

Alexander Erlei

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

2508.18992 2025-09-09 cs.CL cs.AI cs.LG 75%

Automatic Prompt Optimization with Prompt Distillation

Ernest A. Dyagin, Nikita I. Kulin, Artur R. Khairullin, Viktor N. Zhuravlev, Alena N. Sitkina

机构 * Computer Technologies Laboratory（计算机技术实验室）； ITMO University（ITMO大学）

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

2410.08755 2025-09-09 cs.CR cs.MA 75%

PILLAR: an AI-Powered Privacy Threat Modeling Tool

Majid Mollaeefar, Andrea Bissoli, Silvio Ranise

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

2509.06053 2025-09-09 cs.LG cs.AI 73%

PolicyEvolve: Evolving Programmatic Policies by LLMs for multi-player games via Population-Based Training

Mingrui Lv, Hangzhi Liu, Zhi Luo, Hongjie Zhang, Jie Ou

机构 * School of Computer Science（计算机科学学院）； School of Information and Software Engineering（信息与软件工程学院）

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

2509.05449 2025-09-09 cs.LG cs.AI 73%

Neural Breadcrumbs: Membership Inference Attacks on LLMs Through Hidden State and Attention Pattern Analysis

Disha Makhija, Manoj Ghuhan Arivazhagan, Vinayshekhar Bannihatti Kumar, Rashmi Gangadharaiah

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

2503.11132 2025-09-09 cs.CL 70%

X-EcoMLA: Upcycling Pre-Trained Attention into MLA for Efficient and Extreme KV Compression

Guihong Li, Mehdi Rezagholizadeh, Mingyu Yang, Vikram Appia, Emad Barsoum

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

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

2505.17746 2025-09-09 cs.CL 70%

Fast Quiet-STaR: Thinking Without Thought Tokens

Wei Huang, Yizhe Xiong, Xin Ye, Zhijie Deng, Hui Chen, Zijia Lin, Guiguang Ding

机构 * School of Computer Science, Beijing University of Posts and Telecommunications（北京邮电大学计算机学院）； Tsinghua University（清华大学）； Beijing National Research Center for Information Science and Technology (BNRist)（北京信息科学与技术国家研究中心）； Kuaishou Technology（快手科技）； Shanghai Jiao Tong University（上海交通大学）

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

Comments 10 pages, 6 figures

2509.03494 2025-09-09 cs.CV 67%

Parameter-Efficient Adaptation of mPLUG-Owl2 via Pixel-Level Visual Prompts for NR-IQA

Yahya Benmahane, Mohammed El Hassouni

机构 * Computer Science Department Faculty of Sciences, Rabat（科学学院计算机科学系，拉巴特）； Computer Science Department FLSH（计算机科学系FLSH）

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

1. 效率与部署 39 篇

Privacy-Preserving Offloading for Large Language Models in 6G Vehicular Networks

ALPS: Improved Optimization for Highly Sparse One-Shot Pruning for Large Language Models

Fisher Random Walk: Automatic Debiasing Contextual Preference Inference for Large Language Model Evaluation

Efficient Large Language Models with Zero-Shot Adjustable Acceleration

ALPHA: LLM-Enabled Active Learning for Human-Free Network Anomaly Detection

Scaling Intelligence: Designing Data Centers for Next-Gen Language Models

LatticeWorld: A Multimodal Large Language Model-Empowered Framework for Interactive Complex World Generation

ProfilingAgent: Profiling-Guided Agentic Reasoning for Adaptive Model Optimization

Embodied Hazard Mitigation using Vision-Language Models for Autonomous Mobile Robots

Text4Seg++: Advancing Image Segmentation via Generative Language Modeling

The Reel Deal: Designing and Evaluating LLM-Generated Short-Form Educational Videos

Compression Beyond Pixels: Semantic Compression with Multimodal Foundation Models

Human-LLM Synergy in Context-Aware Adaptive Architecture for Scalable Drone Swarm Operation

Guided Decoding and Its Critical Role in Retrieval-Augmented Generation

Crown, Frame, Reverse: Layer-Wise Scaling Variants for LLM Pre-Training

RT-VLM: Re-Thinking Vision Language Model with 4-Clues for Real-World Object Recognition Robustness

Breaking the ICE: Exploring promises and challenges of benchmarks for Inference Carbon & Energy estimation for LLMs

Sensitivity-Aware Post-Training Quantization for Deep Neural Networks

Embedding Poisoning: Bypassing Safety Alignment via Embedding Semantic Shift

Efficient Dynamic Clustering-Based Document Compression for Retrieval-Augmented-Generation

Multimodal Prompt Injection Attacks: Risks and Defenses for Modern LLMs

MaaSO: SLO-aware Orchestration of Heterogeneous Model Instances for MaaS

From Digital Distrust to Codified Honesty: Experimental Evidence on Generative AI in Credence Goods Markets

Automatic Prompt Optimization with Prompt Distillation

PILLAR: an AI-Powered Privacy Threat Modeling Tool

PolicyEvolve: Evolving Programmatic Policies by LLMs for multi-player games via Population-Based Training

Neural Breadcrumbs: Membership Inference Attacks on LLMs Through Hidden State and Attention Pattern Analysis

X-EcoMLA: Upcycling Pre-Trained Attention into MLA for Efficient and Extreme KV Compression

Fast Quiet-STaR: Thinking Without Thought Tokens

Parameter-Efficient Adaptation of mPLUG-Owl2 via Pixel-Level Visual Prompts for NR-IQA