语言大模型 / LLM - arXivDaily 专题

2508.20582 2025-10-13 cs.IR 83%

SUMMA: A Multimodal Large Language Model for Advertisement Summarization

Weitao Jia, Shuo Yin, Zhoufutu Wen, Han Wang, Zehui Dai, Kun Zhang, Zhenyu Li, Tao Zeng, Xiaohui Lv

专题命中指令微调：large language model(title)；language model(title)

2505.18356 2025-10-09 cs.CL cs.AI cs.LG 83%

The Unreasonable Effectiveness of Model Merging for Cross-Lingual Transfer in LLMs

Lucas Bandarkar, Nanyun Peng

机构 * University of California, Los Angeles（加州大学洛杉矶分校）

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

Comments MRL Workshop at EMNLP 2025

2508.10116 2025-08-15 cs.IR 83%

Bridging Modality Gaps in e-Commerce Products via Vision-Language Alignment

Yipeng Zhang, Hongju Yu, Aritra Mandal, Canran Xu, Qunzhi Zhou, Zhe Wu

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

2503.09032 2025-08-08 cs.LG cs.AI cs.CL 83%

Teaching LLMs How to Learn with Contextual Fine-Tuning

Younwoo Choi, Muhammad Adil Asif, Ziwen Han, John Willes, Rahul G. Krishnan

机构 * University of Toronto（多伦多大学）； Vector Institute（向量研究所）

专题命中指令微调：large language model(abstract)；language model(abstract)；instruction tuning(abstract)；prompting(abstract)

Comments ICLR 2025

2406.00222 2025-07-29 cs.CL cs.AI cs.LG 83%

Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training

Maximillian Chen, Ruoxi Sun, Tomas Pfister, Sercan Ö. Arık

机构 * Google（谷歌）； Columbia University（哥伦比亚大学）

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

Comments ICLR 2025; Code: https://github.com/google-research/google-research/tree/master/learning_to_clarify

2507.04664 2025-07-08 cs.CV 83%

VectorLLM: Human-like Extraction of Structured Building Contours vis Multimodal LLMs

Tao Zhang, Shiqing Wei, Shihao Chen, Wenling Yu, Muying Luo, Shunping Ji

机构 * School of Remote Sensing and Information Engineering（遥感与信息工程学院）； College of Oceanography and Space Informatics（海洋学与空间信息学院）； School of Surveying and Geoinformation Engineering（测绘与地理信息工程学院）

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

2506.19767 2025-06-25 cs.CL cs.AI cs.LG 83%

SRFT: A Single-Stage Method with Supervised and Reinforcement Fine-Tuning for Reasoning

Yuqian Fu, Tinghong Chen, Jiajun Chai, Xihuai Wang, Songjun Tu, Guojun Yin, Wei Lin, Qichao Zhang, Yuanheng Zhu, Dongbin Zhao

机构 * Institute of Automation, Chinese Academy of Sciences（中国科学院自动化研究所）； School of Artificial Intelligence, University of Chinese Academy of Sciences（中国科学院大学人工智能学院）； Meituan（美团）； Shanghai Jiao Tong University（上海交通大学）

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

2505.23380 2025-05-30 cs.CV 83%

UniRL: Self-Improving Unified Multimodal Models via Supervised and Reinforcement Learning

Weijia Mao, Zhenheng Yang, Mike Zheng Shou

机构 * Show Lab, National University of Singapore（1 显示实验室，新加坡国立大学）； ByteDance（2 字节跳动）

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

2505.18949 2025-05-27 cs.CL cs.AI cs.LG 83%

The Price of Format: Diversity Collapse in LLMs

Longfei Yun, Chenyang An, Zilong Wang, Letian Peng, Jingbo Shang

机构 * University of California, San Diego（加州大学圣迭戈分校）

专题命中指令微调：large language model(abstract)；language model(abstract)；instruction tuning(abstract)；prompting(abstract)

Comments 14 pages, 7 figures

2407.13419 2025-05-21 cs.CL cs.AI cs.CY cs.LG cs.SC 83%

From Words to Worlds: Compositionality for Cognitive Architectures

Ruchira Dhar, Anders Søgaard

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

Comments Accepted to ICML 2024 Workshop on LLMs & Cognition

2504.12322 2025-04-22 cs.CL cs.AI cs.LG 83%

A Strategic Coordination Framework of Small LLMs Matches Large LLMs in Data Synthesis

Xin Gao, Qizhi Pei, Zinan Tang, Yu Li, Honglin Lin, Jiang Wu, Lijun Wu, Conghui He

机构 * Shanghai AI Laboratory（上海人工智能实验室）； Renmin University of China（中国人民大学）

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

2503.24376 2025-04-01 cs.CV cs.AI cs.CL cs.LG 83%

Exploring the Effect of Reinforcement Learning on Video Understanding: Insights from SEED-Bench-R1

Yi Chen, Yuying Ge, Rui Wang, Yixiao Ge, Lu Qiu, Ying Shan, Xihui Liu

专题命中指令微调：large language model(abstract)；language model(abstract)；post-training(abstract)；SFT(abstract)

Comments Technical Report (In Progress); Code released at: https://github.com/TencentARC/SEED-Bench-R1

2503.07550 2025-03-11 cs.CL cs.AI cs.LG 83%

KSOD: Knowledge Supplement for LLMs On Demand

Haoran Li, Junfeng Hu

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

2410.02320 2025-02-24 cs.CL cs.AI cs.LG 83%

Post-edits Are Preferences Too

Nathaniel Berger, Miriam Exel, Matthias Huck, Stefan Riezler

专题命中指令微调：large language model(abstract)；language model(abstract)；SFT(abstract)；preference optimization(abstract)

Comments To appear at the Ninth Conference on Machine Translation (WMT24)

2411.05194 2024-11-11 cs.LG cs.AI cs.CL 83%

Interactive Dialogue Agents via Reinforcement Learning on Hindsight Regenerations

Joey Hong, Jessica Lin, Anca Dragan, Sergey Levine

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

Comments 23 pages, 5 figures

2411.05048 2024-11-11 cs.CL cs.AI cs.DB cs.IR cs.LG 83%

Leveraging LLMs to Enable Natural Language Search on Go-to-market Platforms

Jesse Yao, Saurav Acharya, Priyaranjan Parida, Srinivas Attipalli, Ali Dasdan

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

Comments 11 pages, 5 figures

2406.11592 2024-10-23 cs.CV 83%

ChildDiffusion: Unlocking the Potential of Generative AI and Controllable Augmentations for Child Facial Data using Stable Diffusion and Large Language Models

Muhammad Ali Farooq, Wang Yao, Peter Corcoran

专题命中指令微调：large language model(title)；language model(title)

Comments This work has been submitted to the IEEE Transactions Journal for possible publication

2405.20974 2024-10-07 cs.CL cs.AI cs.LG 83%

SaySelf: Teaching LLMs to Express Confidence with Self-Reflective Rationales

Tianyang Xu, Shujin Wu, Shizhe Diao, Xiaoze Liu, Xingyao Wang, Yangyi Chen, Jing Gao

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

Comments EMNLP 2024 Main

2409.14713 2024-09-24 cs.CV 83%

Phantom of Latent for Large Language and Vision Models

Byung-Kwan Lee, Sangyun Chung, Chae Won Kim, Beomchan Park, Yong Man Ro

专题命中指令微调：large language model(abstract)；language model(abstract)；instruction tuning(abstract)；SFT(abstract)

Comments Code is available in https://github.com/ByungKwanLee/Phantom

2402.10891 2024-02-19 cs.CL cs.AI cs.LG 83%

Instruction Diversity Drives Generalization To Unseen Tasks

Dylan Zhang, Justin Wang, Francois Charton

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

2311.05584 2023-11-10 cs.LG cs.AI cs.CL 83%

Zero-Shot Goal-Directed Dialogue via RL on Imagined Conversations

Joey Hong, Sergey Levine, Anca Dragan

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

Comments 25 pages, 6 figures

2310.07488 2023-10-20 cs.CL cs.AI cs.LG 83%

KwaiYiiMath: Technical Report

Jiayi Fu, Lei Lin, Xiaoyang Gao, Pengli Liu, Zhengzong Chen, Zhirui Yang, Shengnan Zhang, Xue Zheng, Yan Li, Yuliang Liu, Xucheng Ye, Yiqiao Liao, Chao Liao, Bin Chen, Chengru Song, Junchen Wan, Zijia Lin, Fuzheng Zhang, Zhongyuan Wang, Di Zhang, Kun Gai

专题命中指令微调：large language model(abstract)；language model(abstract)；SFT(abstract)；RLHF(abstract)

Comments technical report. arXiv admin note: text overlap with arXiv:2306.16636 by other authors

2309.07120 2023-09-14 cs.CL cs.AI cs.CV cs.CY cs.LG 83%

Sight Beyond Text: Multi-Modal Training Enhances LLMs in Truthfulness and Ethics

Haoqin Tu, Bingchen Zhao, Chen Wei, Cihang Xie

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

2604.26866 2026-08-18 cs.CL cs.LG 版本更新 83%

MoRFI: Monotonic Sparse Autoencoder Feature Identification

MoRFI: 基于单调性的稀疏自编码器特征识别

Dimitris Dimakopoulos, Shay B. Cohen, Ioannis Konstas

机构 * University of Edinburgh, UK（爱丁堡大学）； Heriot-Watt University, UK（赫瑞斯泰德大学）

专题命中指令微调：language model(abstract,comments)；large language model(abstract)；post-training(abstract)；SFT(abstract)

AI总结本文通过控制微调实验发现，逐步引入新知识会增加幻觉，提出MoRFI方法利用稀疏自编码器分析残差流激活，识别因果相关的潜在特征。

Comments Accepted to the Conference on Language Modeling (COLM) 2026

2304.14999 2023-05-01 cs.CL cs.AI 83%

Empirical Analysis of the Strengths and Weaknesses of PEFT Techniques for LLMs

George Pu, Anirudh Jain, Jihan Yin, Russell Kaplan

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

Comments Short paper, ICLR '23 Workshop on Understanding Foundation Models

2608.15665 2026-08-18 cs.LG 新提交 83%

SubZero+: Efficient Zeroth-Order LLM Fine-Tuning via Large Learning Rates

SubZero+：基于大学习率的高效零阶大语言模型微调方法

Ziming Yu, Shuyao Xiao, Xingyu Zhao, Sike Wang, Pan Zhou, Peiyu Zang, Xiangda Yan, Yongjie Yang, Jia Li

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

AI总结 SubZero+是改进的SubZero框架，通过三种互补方式提升零阶优化稳定性，在1.3B至32B参数模型的SuperGLUE任务上，优于现有零阶基线，扩大稳定学习率范围，缩小与一阶方法差距且内存开销极小。

2608.15311 2026-08-18 cs.AI 新提交 83%

MoE Router-Guided Clustering for Heterogeneous Federated Instruction Tuning

面向异构联邦指令微调的MoE路由器引导聚类

Ankita Sharma, Bahar Farahani, Sanaz Rahimi Moosavi, Amir Rrahmani, Farshad Firouzi, Krishnendu Chakrabarty

机构 * California State University, Dominguez Hills（加州州立大学多明戈斯山分校）； University of California, Irvine（加州大学欧文分校）； Johns Hopkins University（约翰斯·霍普金斯大学）； Arizona State University（亚利桑那州立大学）

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

AI总结该研究提出ClientMorpher框架，通过MoE路由特征设计两种聚类策略，在Databricks Dolly-15K数据集上验证其可提升联邦指令微调的个性化性能，且通信成本与传统方法相当。

2608.11295 2026-08-13 cs.CR cs.AI 新提交 83%

Backdoor Decontamination Dynamics in LLM Agents

大语言模型智能体中的后门净化动力学

Gabriel Huang, Abhay Puri, Léo Boisvert, Alexandre Drouin, Perouz Taslakian, Spandana Gella, Christopher Pal

专题命中指令微调：LLM(title,abstract)；分类 cs.AI

AI总结本研究针对大语言模型智能体的后门问题，提出框架研究净化动力学，发现防御性投毒加遗忘可高效擦除多数后门，中间层仍留触发器感知痕迹。

2608.10657 2026-08-12 eess.IV cs.CV cs.LG 新提交 83%

Retrieval-Augmented Vision Foundation Models for Robust Leukemia Cell Classification across Multiple Microscopy Datasets

用于跨多个显微镜数据集实现稳健白血病细胞分类的检索增强视觉基础模型

Carlos Zamora, Hiram Zuniga, Ulises Orozco-Rosas, Kenia Picos

专题命中指令微调：foundation model(title,abstract)；pretraining(abstract)；分类 cs.LG

AI总结本研究提出基于预训练视觉基础模型的两阶段检索增强框架，经多数据集测试，可实现跨异构数据集的稳健白血病细胞分类，为解决领域偏移问题提供低成本替代方案。

Comments Accepted at SPIE Optics + Photonics 2026 for oral presentation. 23 pages, 12 figures, 9 tables

2605.07284 2026-08-12 cs.LG 版本更新 83%

Same Targets, Different Computation: How Post-Training Divides Work Across Model Layers

指令微调如何改变上游状态影响后期读取：一种跨修补诊断

Yifan Zhou

机构 * University of California, Los Angeles（加州大学洛杉矶分校）

专题命中指令微调：post-training(title,abstract)；SFT(abstract_cn)；分类 cs.LG

AI总结研究探讨指令微调如何通过后期栈与上游状态交互影响后续token选择，通过跨修补诊断发现指令微调模型的后期栈效果依赖自身训练状态，且稀疏特征在其中起中介作用。

Comments 14 pages, 4 figures. Code and compact artifacts: https://github.com/yifan1207/first-divergence-crosspatching

1. 指令微调 11619 篇

SUMMA: A Multimodal Large Language Model for Advertisement Summarization

The Unreasonable Effectiveness of Model Merging for Cross-Lingual Transfer in LLMs

Bridging Modality Gaps in e-Commerce Products via Vision-Language Alignment

Teaching LLMs How to Learn with Contextual Fine-Tuning

Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training

VectorLLM: Human-like Extraction of Structured Building Contours vis Multimodal LLMs

SRFT: A Single-Stage Method with Supervised and Reinforcement Fine-Tuning for Reasoning

UniRL: Self-Improving Unified Multimodal Models via Supervised and Reinforcement Learning

The Price of Format: Diversity Collapse in LLMs

From Words to Worlds: Compositionality for Cognitive Architectures

A Strategic Coordination Framework of Small LLMs Matches Large LLMs in Data Synthesis

Exploring the Effect of Reinforcement Learning on Video Understanding: Insights from SEED-Bench-R1

KSOD: Knowledge Supplement for LLMs On Demand

Post-edits Are Preferences Too

Interactive Dialogue Agents via Reinforcement Learning on Hindsight Regenerations

Leveraging LLMs to Enable Natural Language Search on Go-to-market Platforms

ChildDiffusion: Unlocking the Potential of Generative AI and Controllable Augmentations for Child Facial Data using Stable Diffusion and Large Language Models

SaySelf: Teaching LLMs to Express Confidence with Self-Reflective Rationales

Phantom of Latent for Large Language and Vision Models

Instruction Diversity Drives Generalization To Unseen Tasks

Zero-Shot Goal-Directed Dialogue via RL on Imagined Conversations

KwaiYiiMath: Technical Report

Sight Beyond Text: Multi-Modal Training Enhances LLMs in Truthfulness and Ethics

MoRFI: Monotonic Sparse Autoencoder Feature Identification

Empirical Analysis of the Strengths and Weaknesses of PEFT Techniques for LLMs

SubZero+: Efficient Zeroth-Order LLM Fine-Tuning via Large Learning Rates

MoE Router-Guided Clustering for Heterogeneous Federated Instruction Tuning

Backdoor Decontamination Dynamics in LLM Agents

Retrieval-Augmented Vision Foundation Models for Robust Leukemia Cell Classification across Multiple Microscopy Datasets

Same Targets, Different Computation: How Post-Training Divides Work Across Model Layers