arXivDaily arXiv每日学术速递 周一至周五更新

期刊&会议

International Conference on Machine Learning · 会议 · Machine Learning

共收录 11797
2410.02483 2025-12-09 cs.CV

Event-Customized Image Generation

事件定制图像生成

Zhen Wang, Yilei Jiang, Dong Zheng, Jun Xiao, Long Chen

机构 * Zhejiang University, Hangzhou, China(浙江大学) The Hong Kong University of Science and Technology(香港科技大学)

AI总结 本文提出FreeEvent方法,通过引入实体切换和事件转移路径,实现事件定制化图像生成,提升复杂场景下的定制化能力。

Journal ref ICML 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.06886 2025-12-09 cs.CV

Balanced Learning for Domain Adaptive Semantic Segmentation

领域自适应语义分割中的平衡学习

Wangkai Li, Rui Sun, Bohao Liao, Zhaoyang Li, Tianzhu Zhang

机构 * MoE Key Laboratory of Brain-inspired Intelligent Perception and Cognition(脑启发智能感知与认知关键实验室) University of Science and Technology of China(中国科学技术大学) Deep Space Exploration Laboratory(深空探测实验室)

AI总结 BLDA通过分析logits分布和引入共享锚定分布,有效缓解领域自适应语义分割中的类别偏倚问题,提升模型在欠预测类别上的性能。

Comments Accepted by International Conference on Machine Learning (ICML 2025)

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.03204 2025-12-08 cs.LG

Scaling Internal-State Policy-Gradient Methods for POMDPs

在部分可观测马尔可夫决策过程中的内部状态策略梯度方法扩展

Douglas Aberdeen, Jonathan Baxter

机构 * Research School of Information Science and Engineering, Australian Nat. University, ACT 0200, Australia(信息科学与工程研究学校,澳大利亚国立大学,ACT 0200,澳大利亚) Panscient Pty Ltd, Adelaide, Australia(Panscient Pty Ltd,阿德莱德,澳大利亚)

AI总结 本文提出改进的策略梯度方法,用于在无限时间 horizon 设置中学习具有记忆的策略,通过直接环境模型或模拟解决大规模 POMDPs 问题。

Journal ref Proceedings of the 19th International Conference on Machine Learning (ICML 2002), 2002, pp. 3--10

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.02383 2025-12-08 cs.LG

Reinforcement Learning in POMDP's via Direct Gradient Ascent

通过直接梯度上升在POMDP中进行强化学习

Jonathan Baxter, Peter L. Bartlett

机构 * Research School of Information Sciences and Engineering(信息科学与工程研究学校) Australian National University(澳大利亚国立大学)

AI总结 本文提出GPOMDP算法,通过直接梯度上升在POMDP中优化策略性能,仅需单条样本路径和一个自由参数,证明其收敛性并用于寻找局部最优解。

Journal ref Proceedings of the 17th International Conference on Machine Learning (ICML 2000), 2000, pp. 41--48

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.09625 2025-12-04 cs.LG

GLGENN: A Novel Parameter-Light Equivariant Neural Networks Architecture Based on Clifford Geometric Algebras

GLGENN:一种基于克莱因几何代数的新型参数轻量等变神经网络架构

Ekaterina Filimoshina, Dmitry Shirokov

机构 * HSE University(莫斯科高等经济大学) Skolkovo Institute of Science(斯克尔科沃科学研究院) Institute for Information Transmission Problems of the Russian Academy of Sciences(俄罗斯科学院信息传输问题研究所)

AI总结 GLGENN基于克莱因几何代数提出一种参数轻量的等变神经网络架构,在等变任务中表现优异且参数更少。

Comments ICML 2025, 36 pages

Journal ref Proceedings of the 42nd International Conference on Machine Learning (Vancouver, Canada, 2025), Proceedings of Machine Learning Research, 267, 2025, 17153-17188

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.20260 2025-12-04 cs.LG

Understanding the Limits of Deep Tabular Methods with Temporal Shift

通过时间位移理解深度表格方法的极限

Hao-Run Cai, Han-Jia Ye

机构 * School of Artificial Intelligence, Nanjing University, China(人工智能学院,南京大学,中国) National Key Laboratory for Novel Software Technology, Nanjing University, China(新型软件技术国家重点实验室,南京大学,中国)

AI总结 本文通过改进训练协议和引入时间嵌入方法,提升深度表格模型在时间分布位移下的表现。

Comments 21 pages, 10 figures, 13 tables. ICML 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.03430 2025-12-04 cs.CV

Label-Efficient Hyperspectral Image Classification via Spectral FiLM Modulation of Low-Level Pretrained Diffusion Features

通过低级预训练扩散特征的光谱FiLM调制实现标签高效的超光谱图像分类

Yuzhen Hu, Biplab Banerjee, Saurabh Prasad

机构 * University of Houston, Texas, USA(德克萨斯大学) Indian Institute of Technology Bombay, Mumbai, India(印度班加罗尔理工学院)

AI总结 本文提出了一种基于预训练扩散模型的标签高效超光谱图像分类方法,通过光谱FiLM调制融合空间和光谱信息,提升分类性能。

Comments Accepted to the ICML 2025 TerraBytes Workshop (June 9, 2025)

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.02912 2025-12-03 cs.LG math.ST stat.ML stat.TH

Hypothesis Testing for Generalized Thurstone Models

对广义图斯通模型的假设检验

Anuran Makur, Japneet Singh

机构 * Department of Computer Science, Purdue University, West Lafayette, IN, USA(计算机科学系,普渡大学,西拉法叶,印第安纳州,美国) Elmore Family School of Electrical and Computer Engineering, Purdue University, West Lafayette, IN, USA(埃尔莫尔家族电气与计算机工程学院,普渡大学,西拉法叶,印第安纳州,美国)

AI总结 本文提出了一种针对广义图斯通模型的假设检验方法,通过分离距离分析和反向鞅技术,推导了临界阈值并验证了最小最大下界。

Comments 35 pages, 9 figures

Journal ref 42nd International Conference on Machine Learning (ICML 2025)

详情

展开后加载摘要…

URL PDF HTML 收藏
2512.02633 2025-12-03 cs.AI cs.LG

Zero-Shot Instruction Following in RL via Structured LTL Representations

通过结构化LTL表示实现强化学习中的零样本指令跟随

Mattia Giuri, Mathias Jackermeier, Alessandro Abate

机构 * University of Oxford(牛津大学)

AI总结 本文提出通过结构化LTL表示学习多任务策略,以解决强化学习中多事件交互复杂性问题。

Comments ICML 2025 Workshop on Programmatic Representations for Agent Learning

详情

展开后加载摘要…

URL PDF HTML 收藏
2501.15893 2025-12-03 quant-ph cs.LG

Benchmarking Quantum Reinforcement Learning

量子强化学习的基准测试

Nico Meyer, Christian Ufrecht, George Yammine, Georgios Kontes, Christopher Mutschler, Daniel D. Scherer

机构 * Fraunhofer IIS, Fraunhofer Institute for Integrated Circuits IIS, N\"urnberg, Germany Pattern Recognition Lab, Friedrich-Alexander-Universit\"at Erlangen-N\"urnberg, Erlangen, Germany

AI总结 本文提出了一种新的量子强化学习基准测试方法,通过样本复杂度估计和统计优势定义,评估QRL的性能并质疑其优越性,同时探讨了结果的局限性和对量子优势研究的影响。

Comments Accepted to the 42nd International Conference on Machine Learning (ICML 2025), Vancouver, British Columbia, Canada. 31 pages, 20 figures, 3 tables

Journal ref Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:43934-43964, 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.07255 2025-12-03 cs.AI

Subgoal-Guided Policy Heuristic Search with Learned Subgoals

基于学习子目标的策略树搜索启发式搜索

Jake Tuero, Michael Buro, Levi H. S. Lelis

机构 * Department of Computing Science, University of Alberta, Edmonton, Canada(阿尔伯塔大学计算机科学系) Alberta Machine Intelligence Institute (Amii), Edmonton, Canada(阿尔伯塔机器智能研究所)

AI总结 本文提出了一种基于学习子目标的策略树搜索方法,通过利用搜索过程中生成的树结构来提升策略学习的样本效率。

Comments Accepted to ICML-25

Journal ref Forty-second International Conference on Machine Learning. 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2410.11842 2025-12-02 cs.CV cs.AI cs.LG

MoH: Multi-Head Attention as Mixture-of-Head Attention

MoH:多头注意力作为专家混合注意力

Peng Jin, Bo Zhu, Li Yuan, Shuicheng Yan

机构 * School of Electronic and Computer Engineering, Shenzhen Graduate School, Peking University, Shenzhen, China(电子与计算机工程系,深圳研究生院,北京大学,深圳,中国) Pengcheng Laboratory, Shenzhen, China(鹏城实验室,深圳,中国) School of AI for Science, Shenzhen Graduate School, Peking University, Shenzhen, China(科学人工智能学院,深圳研究生院,北京大学,深圳,中国) National University of Singapore, Singapore(新加坡国立大学,新加坡)

AI总结 MoH通过将注意力头视为专家,提升推理效率并优化性能,仅使用部分注意力头即可超越传统多头注意力。

Comments Accepted by ICML 2025, code: https://github.com/SkyworkAI/MoH

详情

展开后加载摘要…

URL PDF HTML 收藏
2509.10918 2025-12-02 cs.LG cs.AI

ToMA: Token Merge with Attention for Diffusion Models

ToMA: 通过注意力机制的令牌合并用于扩散模型

Wenbo Lu, Shaoyi Zheng, Yuxuan Xia, Shengjie Wang

机构 * Department of Computer Science, New York University(纽约大学计算机科学系)

AI总结 ToMA通过重新设计令牌合并方法,提升扩散模型的GPU效率,减少生成延迟并优化实际性能。

Comments In proceedings of the 42nd International Conference on Machine Learning (ICML 2025). Code available at https://github.com/wenboluu/ToMA

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.10205 2025-12-02 cs.LG

AWP: Activation-Aware Weight Pruning and Quantization with Projected Gradient Descent

AWP: 基于激活感知的权重剪枝与量化方法(投影梯度下降)

Jing Liu, Toshiaki Koike-Akino, Ye Wang, Hassan Mansour, Matthew Brand

机构 * Mitsubishi Electric Research Laboratories (MERL)(三菱电机研究实验室)

AI总结 AWP通过投影梯度下降方法实现激活感知的权重剪枝与量化,优于现有LLM压缩技术。

Comments ICML 2025 workshop on Efficient Systems for Foundation Models

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.19097 2025-12-02 cs.LG stat.ML

Towards Robust Influence Functions with Flat Validation Minima

迈向具有平坦验证极小值的鲁棒影响函数

Xichen Ye, Yifan Wu, Weizhong Zhang, Cheng Jin, Yifan Chen

机构 * Fudan University(复旦大学) Hong Kong Baptist University(香港 Baptist 大学) Shanghai Key Laboratory of Intelligent Information Processing(上海智能信息处理关键实验室) Innovation Center of Calligraphy and Painting Creation Technology(书法和绘画创作技术创新中心)

AI总结 本文提出了一种针对平坦验证极小值的影响函数估计方法,解决了深度学习中因验证风险尖锐性导致的影响估计不准确问题。

Comments Accepted by ICML 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.23220 2025-12-01 cs.CV

Instruction Tuning of Large Language Models for Tabular Data Generation-in One Day

针对表格数据生成的大型语言模型指令微调——一天内完成

Milad Abdollahzadeh, Abdul Raheem, Zilong Zhao, Uzair Javaid, Kevin Yee, Nalam Venkata Abhishek, Tram Truong-Huu, Biplab Sikdar

机构 * SIT(新加坡科技学院) NUS(新加坡国立大学)

AI总结 本文提出了一种在有限资源下通过指令微调提升LLM表格数据生成能力的方法,利用高质量数据集和少量指令实现与GPT-4o相当的生成性能。

Comments Accepted International Conference on Machine Learning (ICML 2025), 1st Workshop on Foundation Models for Structured Data

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.20534 2025-11-26 cs.CL

Bridging the Language Gap: Synthetic Voice Diversity via Latent Mixup for Equitable Speech Recognition

弥合语言鸿沟:通过潜在混合实现合成语音多样性以实现公平的语音识别

Wesley Bian, Xiaofeng Lin, Guang Cheng

机构 * University of California Los Angeles, Department of Statistics(加州大学洛杉矶分校统计学系)

AI总结 本文提出了一种通过潜在混合提升合成语音多样性的方法,以改善低资源语言的语音识别性能。

Comments Accepted at ICML 2025 Workshop on Machine Learning for Audio

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.20531 2025-11-26 cs.AI cs.CV cs.LG

Beyond Generation: Multi-Hop Reasoning for Factual Accuracy in Vision-Language Models

超越生成:面向视觉语言模型事实准确性的多跳推理

Shamima Hossain

机构 * Department of Computer Science, Brac University(布鲁尔大学计算机科学系) bKash Limited(bKash公司)

AI总结 本文提出一种基于知识图谱的多跳推理框架,提升视觉语言模型在事实准确性上的表现,通过多步骤推理增强模型的逻辑推理能力。

Comments Accepted as poster at NewInML Workshop ICML, 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2507.01196 2025-11-26 cs.LG cs.AI cs.ET cs.HC

Are Large Brainwave Foundation Models Capable Yet? Insights from Fine-tuning

大规模脑电基础模型是否已具备能力?来自微调的洞察

Na Lee, Konstantinos Barmpas, Yannis Panagakis, Dimitrios Adamos, Nikolaos Laskaris, Stefanos Zafeiriou

机构 * Imperial College London(伦敦帝国学院) Archimedes / Athena Research Unit(阿基米德/雅典娜研究单位) Aristotle University of Thessaloniki(雅典娜大学) Kapodistrian University of Athens(雅典kapodistrian大学)

AI总结 本文通过微调实验评估了大规模脑电基础模型的能力,发现其在BCI任务中效率有限,提出LoRA技术可提升性能,强调需重新设计架构以提升脑电分析效果。

Journal ref International Conference on Machine Learning (ICML) 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.09341 2025-11-26 cs.AI

Access Controls Will Solve the Dual-Use Dilemma

访问控制将解决双重用途困境

Evžen Wybitul

机构 * ETH Zurich, Switzerland(苏黎世联邦理工学院)

AI总结 本文提出基于访问控制的概念框架,通过验证用户访问双重用途输出,以解决人工智能安全系统在双重用途请求中的决策困境。

Comments Accepted at ICML 2025 Workshop on Technical AI Governance (TAIG)

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.00350 2025-11-25 cs.LG cs.CV

BOOD: Boundary-based Out-Of-Distribution Data Generation

基于边界的Out-Of-Distribution数据生成

Qilin Liao, Shuo Yang, Bo Zhao, Ping Luo, Hengshuang Zhao

机构 * The University of Hong Kong, Hong Kong, China(香港大学) School of AI, Shanghai Jiao Tong University, Shanghai, China(上海交通大学人工智能学院) Department of Computer Science, Harbin Institute of Technology (Shenzhen), Shenzhen, China(哈尔滨工业大学(深圳)计算机科学系)

AI总结 BOOD通过在潜在空间中识别决策边界并生成高质量OOD特征,提升OOD检测性能,实验显示在CIFAR-100数据集上显著优于现有方法。

Comments 14 pages, 8 figures, To be published in the Proceedings of the International Conference on Machine Learning (ICML) 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2504.13837 2025-11-25 cs.AI cs.CL cs.CV

Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?

强化学习真的能激励大语言模型在基础模型之外提升推理能力吗?

Yang Yue, Zhiqi Chen, Rui Lu, Andrew Zhao, Zhaokai Wang, Yang Yue, Shiji Song, Gao Huang

机构 * LeapLab, Tsinghua University(清华大学 LeapLab) Shanghai Jiao Tong University(上海交通大学)

AI总结 本研究发现RLVR方法未有效激发LLMs的新型推理能力,而蒸馏方法能更有效地扩展模型推理能力。

Comments 31 pages, 27 figures

Journal ref NeurIPS 2025 Oral; ICML 2025 AI4MATH workshop best paper

详情

展开后加载摘要…

URL PDF HTML 收藏
2501.17079 2025-11-25 cs.MA cs.AI cs.GT cs.LG

Learning Mean Field Control on Sparse Graphs

在稀疏图上学习均场控制

Christian Fabian, Kai Cui, Heinz Koeppl

机构 * Department of Electrical Engineering(电气工程与信息科技系) Hessian Center for Artificial Intelligence (hessian.AI)(黑森人工智能中心)

AI总结 本文提出了一种针对稀疏图的均场控制模型,通过局部弱收敛理论,设计可扩展的学习算法,以解决多智能体强化学习中稀疏图的挑战。

Comments Accepted at ICML 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.15141 2025-11-21 cs.LG cs.AI stat.ML

BanditSpec: Adaptive Speculative Decoding via Bandit Algorithms

BanditSpec: 通过多臂老虎机算法实现自适应推测解码

Yunlong Hou, Fengzhuo Zhang, Cunxiao Du, Xuan Zhang, Jiachun Pan, Tianyu Pang, Chao Du, Vincent Y. F. Tan, Zhuoran Yang

机构 * National University of Singapore(国立新加坡大学) Sea AI Lab(Sea人工智能实验室) Singapore Management University(新加坡管理学院) Yale University(耶鲁大学)

AI总结 BanditSpec通过多臂老虎机算法实现自适应推测解码,优化超参数选择以提升LLM推理效率。

Comments 35 pages, 4 figures, accepted to ICML, typos and affiliations are corrected

详情

展开后加载摘要…

URL PDF HTML 收藏
2511.16027 2025-11-21 cs.LG cs.AI

HGCN2SP: Hierarchical Graph Convolutional Network for Two-Stage Stochastic Programming

HGCN2SP:用于两阶段随机规划的层次图卷积网络

Yang Wu, Yifan Zhang, Zhenxing Liang, Jian Cheng

机构 * Institute of Automation, Chinese Academy of Sciences, Beijing, China(中国科学院自动化研究所) School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing, China(中国科学院大学人工智能学院) University of Chinese Academy of Sciences, Nanjing, Nanjing, China(中国科学院大学南京校区)

AI总结 HGCN2SP通过层次图卷积网络和强化学习方法,高效解决两阶段随机规划问题,提升求解速度与泛化能力。

Comments 17 pages, 4 figures

Journal ref Proceedings of the 41st International Conference on Machine Learning, PMLR 235:53999-54014, 2024

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.18137 2025-11-20 cs.LG cs.AI cs.CV cs.PF

SpargeAttention: Accurate and Training-free Sparse Attention Accelerating Any Model Inference

Jintao Zhang, Chendong Xiang, Haofeng Huang, Jia Wei, Haocheng Xi, Jun Zhu, Jianfei Chen

机构 * Dept. of Comp. Sci. and Tech., Institute for AI, BNRist Center, THBI Lab, Tsinghua-Bosch Joint ML Center, Tsinghua University(计算机科学与技术系,人工智能研究所,BNRist中心,THBI实验室,清华-博世联合机器学习中心,清华大学) Institute for Interdisciplinary Information Sciences, Tsinghua University(交叉信息学院,清华大学) EECS, University of California, Berkeley(电子工程与计算机科学系,加州大学伯克利分校)

Comments @inproceedings{zhang2025spargeattn, title={Spargeattn: Accurate sparse attention accelerating any model inference}, author={Zhang, Jintao and Xiang, Chendong and Huang, Haofeng and Wei, Jia and Xi, Haocheng and Zhu, Jun and Chen, Jianfei}, booktitle={International Conference on Machine Learning (ICML)}, year={2025} }

Journal ref Proceedings of the 42 nd International Conference on Machine Learning, PMLR 267, 2025 (ICML 2025)

详情

展开后加载摘要…

URL PDF HTML 收藏
2405.10618 2025-11-20 cs.LG math.OC stat.ML

Distributed Event-Based Learning via ADMM

Guner Dilsad Er, Sebastian Trimpe, Michael Muehlebach

机构 * Max Planck Institute for Intelligent Systems(智能系统马克斯·普朗克研究所) Institute for Data Science in Mechanical Engineering(机械工程数据科学研究所)

Comments 35 pages, 12 figures

Journal ref G. Dilsad Er, Sebastian Trimpe, and Michael Muehlebach, Distributed Event-Based Learning via ADMM, International Conference on Machine Learning 267 (2025), pp. 15384-15418

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.22463 2025-11-19 cs.CV cs.LG

Modulated Diffusion: Accelerating Generative Modeling with Modulated Quantization

Weizhi Gao, Zhichao Hou, Junqi Yin, Feiyi Wang, Linyu Peng, Xiaorui Liu

机构 * Department of Computer Science, North Carolina State University(北卡罗来纳州立大学计算机科学系) National Center for Computational Science, Oak Ridge National Lab(橡树岭国家实验室计算科学中心) Department of Mechanical Engineering, Keio University(庆应大学机械工程系)

Comments 26 pages, accepted by ICML 2025

Journal ref Proceedings of the 42nd International Conference on Machine Learning, PMLR 267, 18337-18362, 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2504.16968 2025-11-19 cs.LG cs.AI

BackSlash: Rate Constrained Optimized Training of Large Language Models

Jun Wu, Jiangtao Wen, Yuxing Han

机构 * Shenzhen International Graduate School, Tsinghua University(清华大学深圳国际研究生院) Computer Science, New York University(纽约大学计算机科学)

Journal ref Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:67852-67863, 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.15988 2025-11-19 cs.LG

Near Optimal Decision Trees in a SPLIT Second

Varun Babbar, Hayden McTavish, Cynthia Rudin, Margo Seltzer

机构 * Department of Computer Science, Duke University, Durham, USA(杜克大学计算机科学系) Department of Computer Science, University of British Columbia, Vancouver, Canada(不列颠哥伦比亚大学计算机科学系)

Comments Accepted to ICML 2025 (Oral)

详情

展开后加载摘要…

URL PDF HTML 收藏