语言大模型 / LLM - arXivDaily 专题

2310.05079 2024-03-15 cs.LG 85%

Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?

Cheng Zhang, Jianyi Cheng, Ilia Shumailov, George A. Constantinides, Yiren Zhao

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 by EMNLP2023

2308.14363 2024-03-13 cs.AI 85%

Mobile Foundation Model as Firmware

Jinliang Yuan, Chen Yang, Dongqi Cai, Shihe Wang, Xin Yuan, Zeling Zhang, Xiang Li, Dingge Zhang, Hanzi Mei, Xianqing Jia, Shangguang Wang, Mengwei Xu

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

Comments 17 pages, 15 figures, published to ACM MobiCom'24

Journal ref The 30th Annual International Conference on Mobile Computing and Networking, 2024

2403.01999 2024-03-05 cs.CL 85%

LLM-Oriented Retrieval Tuner

Si Sun, Hanqing Zhang, Zhiyuan Liu, Jie Bao, Dawei Song

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

Comments 16 pages, 8 figures, 5 tables

2402.18649 2024-03-01 cs.CR cs.AI 85%

A New Era in LLM Security: Exploring Security Concerns in Real-World LLM-based Systems

Fangzhou Wu, Ning Zhang, Somesh Jha, Patrick McDaniel, Chaowei Xiao

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

2402.16965 2024-02-28 cs.CR cs.AI 85%

WIPI: A New Web Threat for LLM-Driven Web Agents

Fangzhou Wu, Shutong Wu, Yulong Cao, Chaowei Xiao

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

2303.05352 2024-02-22 cs.CL cond-mat.mtrl-sci 85%

Extracting Accurate Materials Data from Research Papers with Conversational Language Models and Prompt Engineering

Maciej P. Polak, Dane Morgan

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

Comments 7 pages, 3 figures, 1 table

Journal ref Nature Communications (2024) 15:1569

2402.06664 2024-02-19 cs.CR cs.AI 85%

LLM Agents can Autonomously Hack Websites

Richard Fang, Rohan Bindu, Akul Gupta, Qiusi Zhan, Daniel Kang

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

2402.05733 2024-02-09 cs.CL 85%

TimeArena: Shaping Efficient Multitasking Language Agents in a Time-Aware Simulation

Yikai Zhang, Siyu Yuan, Caiyu Hu, Kyle Richardson, Yanghua Xiao, Jiangjie Chen

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

Comments Work in progress

2312.10631 2023-12-19 cs.NI cs.AI 85%

LLM-Twin: Mini-Giant Model-driven Beyond 5G Digital Twin Networking Framework with Semantic Secure Communication and Computation

Yang Hong, Jun Wu, Rosario Morello

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

Comments 18 pages, 11 figures, submitted to Scientific Reports on Nov. 12, 2023

2312.04730 2023-12-14 cs.CR cs.AI 85%

DeceptPrompt: Exploiting LLM-driven Code Generation via Adversarial Natural Language Instructions

Fangzhou Wu, Xiaogeng Liu, Chaowei Xiao

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

2312.03632 2023-12-07 cs.SD cs.LG eess.AS 85%

Multimodal Data and Resource Efficient Device-Directed Speech Detection with Large Foundation Models

Dominik Wagner, Alexander Churchill, Siddharth Sigtia, Panayiotis Georgiou, Matt Mirsamadi, Aarshee Mishra, Erik Marchi

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

2311.14740 2023-11-28 cs.CL 85%

AutoKG: Efficient Automated Knowledge Graph Generation for Language Models

Bohan Chen, Andrea L. Bertozzi

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

Comments 10 pages, accepted by IEEE BigData 2023 as a workshop paper in GTA3

2311.14030 2023-11-27 cs.AI cs.CR 85%

PrivateLoRA For Efficient Privacy Preserving LLM

Yiming Wang, Yu Lin, Xiaodong Zeng, Guannan Zhang

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

2311.06612 2023-11-14 cs.CV cs.CL 85%

PerceptionGPT: Effectively Fusing Visual Perception into LLM

Renjie Pi, Lewei Yao, Jiahui Gao, Jipeng Zhang, Tong Zhang

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

2310.04836 2023-10-10 cs.AI 85%

Dual Grained Quantization: Efficient Fine-Grained Quantization for LLM

Luoming Zhang, Wen Fei, Weijia Wu, Yefei He, Zhenyu Lou, Hong Zhou

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

Comments 15 pages, 2 figures

2308.16369 2023-09-01 cs.LG cs.DC 85%

SARATHI: Efficient LLM Inference by Piggybacking Decodes with Chunked Prefills

Amey Agrawal, Ashish Panwar, Jayashree Mohan, Nipun Kwatra, Bhargav S. Gulavani, Ramachandran Ramjee

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

2308.12908 2023-08-25 cs.DC cs.AR cs.LG 85%

POLCA: Power Oversubscription in LLM Cloud Providers

Pratyush Patel, Esha Choukse, Chaojie Zhang, Íñigo Goiri, Brijesh Warrier, Nithish Mahalingam, Ricardo Bianchini

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

2305.13144 2023-05-30 cs.CL 85%

Response Length Perception and Sequence Scheduling: An LLM-Empowered LLM Inference Pipeline

Zangwei Zheng, Xiaozhe Ren, Fuzhao Xue, Yang Luo, Xin Jiang, Yang You

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

2509.09679 2026-02-12 cs.LG cs.AI cs.CL 85%

ButterflyQuant: Ultra-low-bit LLM Quantization through Learnable Orthogonal Butterfly Transforms

ButterflyQuant: 通过可学习的正交蝴蝶变换实现超低比特LLM量化

Bingxin Xu, Zhen Dong, Oussama Elachqar, Yuzhang Shang

机构 * Bingxin Xu ； Zhen Dong ； Oussama Elachqar ； Yuzhang Shang

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

AI总结 ButterflyQuant通过可学习的正交蝴蝶变换实现超低比特LLM量化，有效抑制异常值并提升量化性能。

Comments Replace discrete Hadamard transforms with continuous Butterfly transforms to facilitate the learning of rotation matrices in LLM quantization

2506.08001 2025-12-12 cs.LG cs.AI cs.CL 85%

Reparameterized LLM Training via Orthogonal Equivalence Transformation

通过正交等价变换重新参数化LLM训练

Zeju Qiu, Simon Buchholz, Tim Z. Xiao, Maximilian Dax, Bernhard Schölkopf, Weiyang Liu

机构 * Max Planck Institute for Intelligent Systems（马克斯·普朗克智能系统研究所）； The Chinese University of Hong Kong（香港中文大学）

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

AI总结 POET通过正交等价变换重新参数化神经元，提升LLM训练的稳定性与泛化能力，并验证了其在大规模神经网络训练中的有效性。

Comments NeurIPS 2025 (40 pages, 26 figures, project page: https://spherelab.ai/poet/, v4: added experiments of finetuning and larger-scale pretraining)

2506.09554 2025-06-13 cs.DC 85%

Understanding the Performance and Power of LLM Inferencing on Edge Accelerators

Mayank Arya, Yogesh Simmhan

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

Comments Extended version of short paper to appear in PAISE 2025: Mayank Arya and Yogesh Simmhan, "Understanding the Performance and Power of LLM Inferencing on Edge Accelerators" in 7th Workshop on Parallel AI and Systems for the Edge, Co-located with IEEE International Parallel & Distributed Processing Symposium (IPDPS), 2025

2407.07845 2024-07-11 cs.GT 85%

Natural Language Mechanisms via Self-Resolution with Foundation Models

Nicolas Della Penna

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

Comments presented as a poster at ACM EC 24 Foundation Models and Game Theory Workshop

2404.11000 2024-05-28 cs.RO 85%

OVAL-Prompt: Open-Vocabulary Affordance Localization for Robot Manipulation through LLM Affordance-Grounding

Edmond Tong, Anthony Opipari, Stanley Lewis, Zhen Zeng, Odest Chadwicke Jenkins

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

Comments Accepted to Vision-Language Models for Navigation and Manipulation (VLMNM) Workshop (ICRA 2024)

2608.23018 2026-08-25 cs.LG cs.AI 新提交 85%

SplitLite: Low-Rank Residual Compression for Split Learning

SplitLite：面向分裂学习的低秩残差压缩

Tao Li, Yulin Tang, Qi Guo, Xianhao Chen

机构 * The University of Hong Kong（香港大学）； California Institute of Technology（加州理工学院）

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

AI总结本文针对设备端LLM联邦微调的高通信成本问题，提出SplitLite方法，利用残差的低秩结构压缩传输量，在保持性能的同时大幅降低了通信成本。

2608.20390 2026-08-24 cs.CL cs.AI cs.CY 新提交 85%

Ansari: A Retrieval-Grounded Islamic AI Assistant -- Architecture, Deployment, and Lessons from 140,000 Conversations

Ansari：基于检索的伊斯兰AI助手——架构、部署及14万次对话的经验教训

M Waleed Kadous, Amr Elsayed, Abdullah Al Nahas, Ashraf Haress

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

AI总结本研究推出基于检索的伊斯兰AI助手Ansari，其经多平台部署后处理14万次跨语言对话，在IslamicMMLU等基准中表现优异，为信仰敏感型LLM部署提供了经验。

Comments 10 pages, 1 figure, 3 tables. Live system: https://askansari.ai . Code: https://github.com/ansari-project

2603.02830 2026-08-18 cs.CL cs.AI 版本更新 85%

Faster, Cheaper, More Accurate: Specialised Knowledge Tracing Models Outperform LLMs

更快、更便宜、更准确：专门知识追踪模型超越大语言模型

Prarthana Bhattacharyya, Joshua Mitton, Ralph Abboud, Simon Woodhead

机构 * Prarthana Bhattacharyya（普拉塔拉·巴特查利亚）； Joshua Mitton（乔舒亚·米顿）； Ralph Abboud（拉尔夫·阿布德）； Simon Woodhead（西蒙·伍德赫德）

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

AI总结本文比较了多种LLM和KT模型在预测性能、部署成本和推理速度上的表现，发现领域特定的KT模型在准确性上优于LLM。

Comments Published as Poster at EDM 2026. Link: https://educationaldatamining.org/edm2026/proceedings/2026.EDM.poster-demo-papers.379/2026.EDM.poster-demo-papers.379.pdf. 7 pages, 6 figures. Prarthana Bhattacharyya and Joshua Mitton contributed equally to this work

2608.10939 2026-08-12 cs.CL cs.AI 新提交 85%

A Cost-Efficient Routing Pipeline for Multilingual Short-Text Classification Using Small Language Models

使用小型语言模型的低成本多语言短文本分类路由流水线

Wajdi Ben Saad, Safa Madiouni

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

AI总结本研究提出一种使用小型语言模型的多语言短文本分类路由流水线，通过选择性翻译低资源语言提升分类性能，在两个基准上验证了该策略的有效性。

Comments Accepted for publication at the 16th International Conference on Advanced Computer Information Technologies (ACIT 2026), https://acit.tech/

2607.29602 2026-08-03 cs.CL cs.AI cs.CV cs.HC 新提交 85%

FriendBench: Benchmarking Dyadic Familiarity Inference in Humans and Multimodal Large Language Models

FriendBench：人类与多模态大语言模型的二元熟悉度推理基准

Jeffrey M. Girard, Jason Z. Zheng, Jacqueline R. Vertino, Antony D'Avirro, Benjamin Peloquin

机构 * Fluid Concepts Research（流体概念研究机构）

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

AI总结该研究推出FriendBench基准，通过20秒二元破冰对话片段推断两人熟悉度，对比7家公司26个模型与人类的表现，发现模型与人类准确率无统计差异但先验倾向不同，且仅人类能从可见行为中获益，同时发布了相关资源。

Comments 15 pages, 3 figures

2607.28319 2026-07-31 cs.CL cs.CY cs.LG 新提交 85%

Fairness Pruning: Locating Demographic Bias in GLU-MLP Layers via Differential Activations

公平剪枝：通过差分激活定位GLU-MLP层中的人口统计偏差

Pere Martra, Eugenio Martínez Cámara, Alfonso Ureña López

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

AI总结本研究提出Fairness Pruning方法，通过最小对比提示对与激活捕获定位LLM的GLU-MLP层中人口统计偏差神经元，经实验验证该方法可针对性调控偏差且对模型能力影响极小。

Comments 15 pages, 3 figures, 9 tables. Code and datasets publicly available

2501.02211 2026-07-24 cs.CV cs.CL cs.LG 版本更新 85%

How Robust Is Homogeneity Bias in LLMs? Evidence Across Models, Decoding Settings, and Identity Signals

开放权重大语言模型中的同质性偏见对解码超参数具有鲁棒性

Messi H. J. Lee

机构 * Independent Researcher（独立研究者）

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

AI总结研究开放权重LLM中同质性偏见对解码超参数的鲁棒性，发现西班牙裔和亚裔美国人的同质性偏见在大多数配置下稳健，而非裔美国人和性别偏见则因模型而异。

Comments 19 pages

1. 效率与部署 22368 篇

Revisiting Block-based Quantisation: What is Important for Sub-8-bit LLM Inference?

Mobile Foundation Model as Firmware

LLM-Oriented Retrieval Tuner

A New Era in LLM Security: Exploring Security Concerns in Real-World LLM-based Systems

WIPI: A New Web Threat for LLM-Driven Web Agents

Extracting Accurate Materials Data from Research Papers with Conversational Language Models and Prompt Engineering

LLM Agents can Autonomously Hack Websites

TimeArena: Shaping Efficient Multitasking Language Agents in a Time-Aware Simulation

LLM-Twin: Mini-Giant Model-driven Beyond 5G Digital Twin Networking Framework with Semantic Secure Communication and Computation

DeceptPrompt: Exploiting LLM-driven Code Generation via Adversarial Natural Language Instructions

Multimodal Data and Resource Efficient Device-Directed Speech Detection with Large Foundation Models

AutoKG: Efficient Automated Knowledge Graph Generation for Language Models

PrivateLoRA For Efficient Privacy Preserving LLM

PerceptionGPT: Effectively Fusing Visual Perception into LLM

Dual Grained Quantization: Efficient Fine-Grained Quantization for LLM

SARATHI: Efficient LLM Inference by Piggybacking Decodes with Chunked Prefills

POLCA: Power Oversubscription in LLM Cloud Providers

Response Length Perception and Sequence Scheduling: An LLM-Empowered LLM Inference Pipeline

ButterflyQuant: Ultra-low-bit LLM Quantization through Learnable Orthogonal Butterfly Transforms

Reparameterized LLM Training via Orthogonal Equivalence Transformation

Understanding the Performance and Power of LLM Inferencing on Edge Accelerators

Natural Language Mechanisms via Self-Resolution with Foundation Models

OVAL-Prompt: Open-Vocabulary Affordance Localization for Robot Manipulation through LLM Affordance-Grounding

SplitLite: Low-Rank Residual Compression for Split Learning

Ansari: A Retrieval-Grounded Islamic AI Assistant -- Architecture, Deployment, and Lessons from 140,000 Conversations

Faster, Cheaper, More Accurate: Specialised Knowledge Tracing Models Outperform LLMs

A Cost-Efficient Routing Pipeline for Multilingual Short-Text Classification Using Small Language Models

FriendBench: Benchmarking Dyadic Familiarity Inference in Humans and Multimodal Large Language Models

Fairness Pruning: Locating Demographic Bias in GLU-MLP Layers via Differential Activations

How Robust Is Homogeneity Bias in LLMs? Evidence Across Models, Decoding Settings, and Identity Signals