语言大模型 / LLM - arXivDaily 专题

2603.04707 2026-03-06 cs.CL cs.AI 88%

Detection of Illicit Content on Online Marketplaces using Large Language Models

利用大语言模型检测在线市场上的非法内容

Quoc Khoa Tran, Thanh Thi Nguyen, Campbell Wilson

机构 * AiLECS Lab Monash University（AiLECS实验室莫纳什大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

AI总结本文研究了利用大语言模型检测在线市场非法内容的有效性，发现LLM在复杂分类任务中表现优于传统方法。

Comments Accepted for publication in the Proceedings of the 8th International Conference on Natural Language Processing (ICNLP 2026)

2603.03313 2026-03-05 cs.CL cs.AI 88%

How does fine-tuning improve sensorimotor representations in large language models?

微调如何改善大语言模型中的感知运动表征

Minghua Wu, Javier Conde, Pedro Reviriego, Marc Brysbaert

机构 * Department of Experimental Psychology, Ghent University（实验心理学系，根特大学）； Information Processing and Telecommunications Center (IPTC), Universidad Politécnica de Madrid, Spain（信息处理与电信中心（IPTC），马德里理工大学，西班牙）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

AI总结本研究通过微调改善大语言模型的感知运动表征，发现其在语言和感知运动维度上具有良好的泛化能力，但对学习目标敏感，无法跨任务格式迁移。

2506.11103 2026-03-04 cs.CL cs.AI 88%

You Only Fine-tune Once: Many-Shot In-Context Fine-Tuning for Large Language Models

你只需微调一次：用于大型语言模型的多示例上下文微调

Wenchong He, Liqian Peng, Zhe Jiang, Alex Go

机构 * University of Florida（佛罗里达大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

AI总结本文提出多示例上下文微调方法，通过扩展上下文学习原理，显著提升大型语言模型在多任务处理中的性能，接近专门微调水平并缓解灾难性遗忘问题。

Comments 20 pages, 6 figures, 12 tables

2602.21223 2026-02-26 cs.CL cs.AI 88%

Measuring Pragmatic Influence in Large Language Model Instructions

在大型语言模型指令中测量隐含影响

Yilin Geng, Omri Abend, Eduard Hovy, Lea Frermann

机构 * University of Melbourne（墨尔本大学）； Hebrew University of Jerusalem（耶路撒冷希伯来大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

AI总结本文提出了一种框架分解和分类方法，系统测量大型语言模型指令中的隐含影响，揭示了框架对指令优先级的可预测影响。

2506.06060 2026-02-26 cs.CL cs.AI 88%

Simple Yet Effective: Extracting Private Data Across Clients in Federated Fine-Tuning of Large Language Models

简单而有效的：在联邦微调大型语言模型中提取客户端私有数据

Yingqi Hu, Zhuo Zhang, Jingyuan Zhang, Jinghua Wang, Qifan Wang, Lizhen Qu, Zenglin Xu

机构 * Harbin Institute of Technology（哈尔滨工业大学）； Kuaishou Technology（快手科技）； Meta AI ； Monash University（墨尔本大学）； Fudan University（复旦大学）； Shanghai Academy of Artificial Intelligence for Science（上海人工智能科学研究院）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

AI总结本文提出三种简单有效的方法，用于在联邦微调大型语言模型中提取客户端私有数据，揭示了FedLLMs中的隐私风险并建立了评估框架。

Comments IJCNLP 2025 Findings

2602.15449 2026-02-18 cs.CL cs.LG cs.SE 88%

TAROT: Test-driven and Capability-adaptive Curriculum Reinforcement Fine-tuning for Code Generation with Large Language Models

TAROT: 为基于大语言模型的代码生成设计的测试驱动和能力适应课程强化微调

Chansung Park, Juyong Jiang, Fan Wang, Sayak Paul, Jiasi Shen, Jing Tang, Jianguo Li

机构 * Electronics and Telecommunications Research Institute（电信研究所）； The Hong Kong University of Science and Technology (Guangzhou)（香港科技大学（广州））； The Hong Kong University of Science and Technology（香港科技大学）； Hugging Face ； Ant Group（蚂蚁集团）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.LG

AI总结 TAROT通过测试驱动和能力适应的课程强化微调方法，提升大语言模型生成代码的功能正确性和稳健性。

Comments The first three authors contributed equally to this work; listing order is random

2602.05148 2026-02-13 cs.LG cs.AI 88%

CoSA: Compressed Sensing-Based Adaptation of Large Language Models

CoSA: 基于压缩感知的大型语言模型适应

Songtao Wei, Yi Li, Bohan Zhang, Zhichun Guo, Ying Huang, Yuede Ji, Miao Yin, Guanpeng Li, Bingzhe Li

机构 * Department of Computer Science, University of Texas at Dallas, Richardson, United States（计算机科学系，德克萨斯大学达拉斯分校，理查德森，美国）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI、cs.LG

AI总结 CoSA基于压缩感知理论提出了一种新的参数高效微调方法，通过固定随机投影矩阵和紧凑可学习核心实现高效且表达性强的多尺度模型适应。

2502.13313 2026-02-10 cs.AI cs.LG 88%

Revisiting Privacy, Utility, and Efficiency Trade-offs when Fine-Tuning Large Language Models

重新审视在微调大语言模型时隐私、效用和效率之间的权衡

Soumi Das, Camila Kolling, Mohammad Aflah Khan, Mahsa Amani, Bishwamittra Ghosh, Qinyuan Wu, Till Speicher, Krishna P. Gummadi

机构 * MPI-SWS Germany（德国马克斯·普朗克研究所-斯图加特）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI、cs.LG

AI总结本文研究了在微调大语言模型时隐私、效用和效率之间的权衡，发现高效微调方法如LoRA在缓解隐私风险方面与差分隐私方法如DP效果相当，挑战了传统观点。

Comments This work has been accepted at IASEAI 2026 (Non-archival)

2602.07798 2026-02-10 cs.LG cs.AI 88%

CausalTAD: Injecting Causal Knowledge into Large Language Models for Tabular Anomaly Detection

CausalTAD: 在表格异常检测中向大语言模型注入因果知识

Ruiqi Wang, Ruikang Liu, Runyu Chen, Haoxiang Suo, Zhiyi Peng, Zhuo Tang, Changjian Chen

机构 * College of Computer Science and Electronic Engineering, Hunan University, China（计算机科学与电子工程学院，湖南大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI、cs.LG

AI总结 CausalTAD通过注入因果知识提升表格异常检测性能，采用因果关系重构与加权策略，实验表明优于现有方法。

2602.04398 2026-02-05 cs.CL cs.AI 88%

Bi-directional Bias Attribution: Debiasing Large Language Models without Modifying Prompts

双向偏见归因：无需修改提示的大型语言模型去偏方法

Yujie Lin, Kunquan Li, Yixuan Liao, Xiaoxin Chen, Jinsong Su

机构 * School of Informatics, Xiamen University, China（厦门大学信息学院）； vivo AI Lab, China（vivo AI实验室）； Key Laboratory of Digital Protection and Intelligent Processing of Intangible Cultural Heritage of Fujian and Taiwan (Xiamen University), Ministry of Culture and Tourism, China（福建省和台湾非物质文化遗产数字化保护与智能处理重点实验室（厦门大学），文化部，中国）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

AI总结本文提出一种无需微调或修改提示的双向偏见归因方法，通过检测刻板印象词汇并归因神经元偏见，有效减少LLM中的偏见同时保持性能。

2509.25736 2026-02-02 cs.CL cs.AI cs.IT cs.NI math.IT 88%

Think Less, Label Better: Multi-Stage Domain-Grounded Synthetic Data Generation for Fine-Tuning Large Language Models in Telecommunications

少思多标：多阶段领域导向的合成数据生成用于电信领域大型语言模型微调

Chenhua Shi, Gregor Macdonald, Bhavika Jalli, Wanlu Lei, John Zou, Mridul Jain, Joji Philip

机构 * Ericsson（爱立信）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

AI总结本文提出一种多阶段自动化方法，利用领域知识图生成高质量合成数据，用于电信领域大型语言模型的微调，减少人工标注依赖，提升数据质量。

Comments 6 pages, 6 figures, 5 tables, IEEE ICC 2026

2509.21044 2026-01-28 cs.LG cs.AI 88%

Reinforcement Learning Fine-Tuning Enhances Activation Intensity and Diversity in the Internal Circuitry of LLMs

强化学习微调增强大语言模型内部电路的激活强度和多样性

Honglin Zhang, Qianyue Hao, Fengli Xu, Yong Li

机构 * Department of Electronic Engineering, BNRist, Tsinghua University（电子工程系，BNRist，清华大学）

专题命中指令微调：LLM(abstract)；large language model(abstract)；language model(abstract)；pretraining(abstract)

AI总结强化学习微调增强了LLMs内部电路的激活强度和多样性，相较于偏好优化方法，其在数学泛化上的优势源于信息流的冗余与灵活性。

2601.17533 2026-01-27 cs.CR cs.AI cs.LG 88%

Reconstructing Training Data from Adapter-based Federated Large Language Models

从基于适配器的联邦大语言模型中重建训练数据

Silong Chen, Yuchuan Luo, Guilin Deng, Yi Liu, Min Xu, Shaojing Fu, Xiaohua Jia

机构 * National University of Defense Technology（国防科技大学）； City University of Hong Kong（香港城市大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI、cs.LG

AI总结研究提出UTR攻击，揭示基于适配器的FedLLMs中低秩适配器导致的隐私泄露问题，挑战轻量级适应提升安全性的假设。

Comments Yuchuan Luo and Yi Liu are co-corresponding authors. Accepted at The Web Conference (WWW) 2026

2601.06788 2026-01-27 cs.LG cs.AI hep-th quant-ph stat.ML 88%

Artificial Entanglement in the Fine-Tuning of Large Language Models

大语言模型微调中的人工纠缠

Min Chen, Zihan Wang, Canyu Chen, Zeguan Wu, Manling Li, Junyu Liu

机构 * Department of Computer Science, University of Pittsburgh, Pittsburgh, PA 15260, USA（匹兹堡大学计算机科学系）； Department of Computer Science, Northwestern University, Evanston, IL 60208, USA（西北大学计算机科学系）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI、cs.LG

AI总结本研究通过量子信息视角分析大语言模型微调中的人工纠缠，揭示LoRA和FFT在参数结构上的差异，但这种差异不影响注意力输出，提出低秩更新的有效性源于'无毛定理'特性。

Comments 41 pages, many figures

2503.01839 2026-01-09 cs.CR cs.AI cs.CL cs.CV 88%

Jailbreaking Safeguarded Text-to-Image Models via Large Language Models

通过大型语言模型对受保护的文本到图像模型进行劫持

Zhengyuan Jiang, Yuepeng Hu, Yuchen Yang, Yinzhi Cao, Neil Zhenqiang Gong

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

AI总结本文提出了一种利用微调大型语言模型来劫持受安全防护的文本到图像模型的方法，有效绕过安全防护并优于现有攻击技术。

Comments Accepted by EACL 2026 Findings

2504.16628 2026-01-01 cs.LG cs.CL 88%

ParetoHqD: Fast Offline Multiobjective Alignment of Large Language Models using Pareto High-quality Data

ParetoHqD: 利用帕累托高质量数据快速实现大语言模型的多目标对齐

Haoran Gu, Handing Wang, Yi Mei, Mengjie Zhang, Yaochu Jin

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.LG

AI总结 ParetoHqD通过利用帕累托高质量数据实现大语言模型的多目标对齐，提升对齐效果和效率。

Comments Accepted as a main conference paper at AAAI 2026

2512.20633 2025-12-25 cs.LG cs.AI 88%

Enhancing Lung Cancer Treatment Outcome Prediction through Semantic Feature Engineering Using Large Language Models

通过大语言模型进行语义特征工程提升肺癌治疗预后预测

MunHwan Lee, Shaika Chowdhury, Xiaodi Li, Sivaraman Rajaganapathy, Eric W Klee, Ping Yang, Terence Sio, Liewei Wang, James Cerhan, Nansu NA Zong

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI、cs.LG

AI总结本研究利用大语言模型进行语义特征工程，提升肺癌治疗预后预测的准确性与可解释性。

2512.10150 2025-12-12 cs.CL cs.AI 88%

Unforgotten Safety: Preserving Safety Alignment of Large Language Models with Continual Learning

遗忘的安全性：通过持续学习保持大型语言模型的安全对齐

Lama Alssum, Hani Itani, Hasan Abed Al Kader Hammoud, Philip Torr, Adel Bibi, Bernard Ghanem

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

AI总结本文通过持续学习方法有效缓解大型语言模型在适应新任务时的安全退化问题，证明了持续学习在保持模型安全性和任务实用性方面的有效性。

2512.01326 2025-12-02 cs.CR cs.CL cs.LG 88%

Securing Large Language Models (LLMs) from Prompt Injection Attacks

保护大型语言模型（LLMs）免受提示注入攻击

Omar Farooq Khan Suri, John McCrae

机构 * University of Galway, Ireland（Galway大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.LG

AI总结本研究评估了JATMO方法对提示注入攻击的鲁棒性，发现其在减少攻击成功率的同时仍存在漏洞，提示需要更复杂的防御策略。

Comments 10 pages, 1 figure, 1 table

2511.21733 2025-12-01 cs.CL cs.AI 88%

RoSA: Enhancing Parameter-Efficient Fine-Tuning via RoPE-aware Selective Adaptation in Large Language Models

RoSA: 通过RoPE-aware选择性适应提升大语言模型的参数高效微调

Dayan Pan, Jingyuan Wang, Yilong Zhou, Jiawei Cheng, Pengyue Jia, Xiangyu Zhao

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

AI总结 RoSA通过RoPE-aware选择性适应提升大语言模型的参数高效微调效率，结合注意力增强与动态层选择策略，实现更精准的模型适应。

Comments Accepted by AAAI' 26

2411.02462 2025-11-25 cs.SE cs.AI cs.LG 88%

Parameter-Efficient Fine-Tuning of Large Language Models for Unit Test Generation: An Empirical Study

大型语言模型参数高效微调用于单元测试生成：一项实证研究

André Storhaug, Jingyue Li

机构 * Norwegian University of Science and Technology（挪威科学技术大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI、cs.LG

AI总结本文通过实证研究探讨了大型语言模型在单元测试生成中的参数高效微调方法，发现LoRA在某些情况下性能可与全量微调媲美，提示微调在成本效益上表现最佳。

Comments 26 pages, 2 figures, 6 tables, 1 listing

2509.13625 2025-11-20 cs.LG cs.CL cs.CR 88%

Privacy Preserving In-Context-Learning Framework for Large Language Models

Bishnu Bhusal, Manoj Acharya, Ramneet Kaur, Colin Samplawski, Anirban Roy, Adam D. Cobb, Rohit Chadha, Susmit Jha

机构 * SRI

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.LG

Comments Git repo: https://github.com/bhusalb/privacy-preserving-icl

2511.08968 2025-11-13 cs.LG cs.CL 88%

Bayesian Mixture of Experts For Large Language Models

Maryam Dialameh, Hossein Rajabzadeh, Weiwei Zhang, Walid Ahmed, Hyock Ju Kwon

机构 * University of Waterloo（滑铁卢大学）； Ascend Team, Huawei Technologies（华为技术有限公司）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.LG

2511.08597 2025-11-13 cs.CL cs.AI 88%

Self-HarmLLM: Can Large Language Model Harm Itself?

Heehwan Kim, Sungjune Park, Daeseon Choi

机构 * Soongsil University（首尔大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

2511.07581 2025-11-12 cs.AI cs.CL cs.IR 88%

Think Before You Retrieve: Learning Test-Time Adaptive Search with Small Language Models

Supriti Vijay, Aman Priyanshu, Anu Vellore, Baturay Saglam, Amin Karbasi

机构 * Carnegie Mellon University（卡内基梅隆大学）； Foundation AI–Cisco Systems Inc.（Foundation AI–Cisco 系统公司）； Yale University（耶鲁大学）

专题命中指令微调：language model(title,abstract)；small language model(title)；large language model(abstract)；分类 cs.CL、cs.AI

Comments 37 images, 7 figures, and 15 tables

2511.03945 2025-11-07 cs.CL cs.AI 88%

Direct Semantic Communication Between Large Language Models via Vector Translation

Fu-Chun Yang, Jason Eshraghian

机构 * University of California, Santa Cruz（加州大学圣克ruz分校）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

Comments 9 pages, 1 figure, 2 tables

2502.11559 2025-11-04 cs.CL cs.AI 88%

Auto-Search and Refinement: An Automated Framework for Gender Bias Mitigation in Large Language Models

Yue Xu, Chengyan Fu, Li Xiong, Sibei Yang, Wenjie Wang

机构 * School of Information Science and Technology, ShanghaiTech University（信息科学与技术学院，上海科技大学）； Emory University（埃默里大学）； Sun Yat-sen University（中山大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

Comments Accepted to NeurIPS 2025

2510.27172 2025-11-03 cs.LG cs.AI 88%

Adaptive Defense against Harmful Fine-Tuning for Large Language Models via Bayesian Data Scheduler

Zixuan Hu, Li Shen, Zhenyi Wang, Yongxian Wei, Dacheng Tao

机构 * Nanyang Technological University（南洋理工大学）； Shenzhen Campus of Sun Yat-sen University（中山大学深圳校区）； University of Central Florida（佛罗里达大学）； Tsinghua University（清华大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI、cs.LG

2507.13949 2025-10-23 cs.CL cs.AI 88%

Exploiting Primacy Effect To Improve Large Language Models

Bianca Raimondi, Maurizio Gabbrielli

机构 * University of Bologna（博洛尼亚大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.CL、cs.AI

Comments Accepted by RANLP 2025

2510.16089 2025-10-21 cs.LG cs.AI 88%

STABLE: Gated Continual Learning for Large Language Models

William Hoy, Nurcin Celik

机构 * Department of Industrial and Systems Engineering, University of Miami（工业与系统工程系，迈阿密大学）

专题命中指令微调：large language model(title,abstract)；language model(title,abstract)；分类 cs.AI、cs.LG

1. 指令微调 11567 篇

Detection of Illicit Content on Online Marketplaces using Large Language Models

How does fine-tuning improve sensorimotor representations in large language models?

You Only Fine-tune Once: Many-Shot In-Context Fine-Tuning for Large Language Models

Measuring Pragmatic Influence in Large Language Model Instructions

Simple Yet Effective: Extracting Private Data Across Clients in Federated Fine-Tuning of Large Language Models

TAROT: Test-driven and Capability-adaptive Curriculum Reinforcement Fine-tuning for Code Generation with Large Language Models

CoSA: Compressed Sensing-Based Adaptation of Large Language Models

Revisiting Privacy, Utility, and Efficiency Trade-offs when Fine-Tuning Large Language Models

CausalTAD: Injecting Causal Knowledge into Large Language Models for Tabular Anomaly Detection

Bi-directional Bias Attribution: Debiasing Large Language Models without Modifying Prompts

Think Less, Label Better: Multi-Stage Domain-Grounded Synthetic Data Generation for Fine-Tuning Large Language Models in Telecommunications

Reinforcement Learning Fine-Tuning Enhances Activation Intensity and Diversity in the Internal Circuitry of LLMs

Reconstructing Training Data from Adapter-based Federated Large Language Models

Artificial Entanglement in the Fine-Tuning of Large Language Models

Jailbreaking Safeguarded Text-to-Image Models via Large Language Models

ParetoHqD: Fast Offline Multiobjective Alignment of Large Language Models using Pareto High-quality Data

Enhancing Lung Cancer Treatment Outcome Prediction through Semantic Feature Engineering Using Large Language Models

Unforgotten Safety: Preserving Safety Alignment of Large Language Models with Continual Learning

Securing Large Language Models (LLMs) from Prompt Injection Attacks

RoSA: Enhancing Parameter-Efficient Fine-Tuning via RoPE-aware Selective Adaptation in Large Language Models

Parameter-Efficient Fine-Tuning of Large Language Models for Unit Test Generation: An Empirical Study

Privacy Preserving In-Context-Learning Framework for Large Language Models

Bayesian Mixture of Experts For Large Language Models

Self-HarmLLM: Can Large Language Model Harm Itself?

Think Before You Retrieve: Learning Test-Time Adaptive Search with Small Language Models

Direct Semantic Communication Between Large Language Models via Vector Translation

Auto-Search and Refinement: An Automated Framework for Gender Bias Mitigation in Large Language Models

Adaptive Defense against Harmful Fine-Tuning for Large Language Models via Bayesian Data Scheduler

Exploiting Primacy Effect To Improve Large Language Models

STABLE: Gated Continual Learning for Large Language Models