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

2026-02-03 至 2026-02-03 共收录 7
2602.01708 2026-02-03 cs.CL cs.AI cs.GT

Game of Thought: Robust Information Seeking with Large Language Models Using Game Theory

思维博弈:利用博弈论的大型语言模型鲁棒信息检索

Langyuan Cui, Chun Kai Ling, Hwee Tou Ng

机构 * Department of Computer Science, National University of Singapore, 13 Computing Drive, Singapore 117417(新加坡国立大学计算机科学系)

AI总结 本文提出Game of Thought框架,通过博弈论技术提升大型语言模型在信息检索中的鲁棒性,实验证明其在最坏情况下的性能优于传统方法。

Comments 23 pages, 10 figures, under review at ICML 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2106.14568 2026-02-03 cs.LG cs.CV

Deep Ensembling with No Overhead for either Training or Testing: The All-Round Blessings of Dynamic Sparsity

深度集成无额外开销:动态稀疏性对训练和测试的全方位祝福

Shiwei Liu, Tianlong Chen, Zahra Atashgahi, Xiaohan Chen, Ghada Sokar, Elena Mocanu, Mykola Pechenizkiy, Zhangyang Wang, Decebal Constantin Mocanu

机构 * Eindhoven University of Technology(埃因霍温理工大学) University of Texas at Austin(德克萨斯大学奥斯汀分校) University of Twente(特文特大学)

AI总结 FreeTickets通过动态稀疏训练实现高效集成,以更低的计算成本提升预测精度和鲁棒性。

Comments published in International Conference on Learning Representations (ICLR 2022)

Journal ref Proceedings of the International Conference on Machine Learning (ICLR 2022)

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.01237 2026-02-03 cs.AI

Predictive Scheduling for Efficient Inference-Time Reasoning in Large Language Models

为大语言模型的高效推理时间推理进行预测调度

Katrina Brown, Aneesh Muppidi, Rana Shahout

机构 * Harvard College(哈佛学院) Harvard SEAS(哈佛工程与应用科学学院)

AI总结 预测调度通过预运行轻量级预测器优化标记预算分配,提升大语言模型在推理任务中的准确性和效率。

Comments ICML ES-FoMo 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.00869 2026-02-03 cs.LG cs.AI cs.NA math.NA

Improving Flow Matching by Aligning Flow Divergence

通过对齐流发散性来改进流匹配

Yuhao Huang, Taos Transue, Shih-Hsin Wang, William Feldman, Hong Zhang, Bao Wang

机构 * Department of Mathematics, University of Utah, Salt Lake City, UT, USA(犹他大学数学系) Imaging (SCI) Institute, Salt Lake City, UT, USA(成像(SCI)研究所) Computer Science Division, 240 Argonne National Laboratory, Lemont, IL, USA(计算机科学部,阿贡国家实验室)

AI总结 本文提出了一种新的流匹配方法,通过同时匹配流及其发散性来提升生成模型的性能。

Comments Published in ICML 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2509.25741 2026-02-03 stat.ML cs.LG

Test time training enhances in-context learning of nonlinear functions

测试时间训练增强非线性函数的上下文学习

Kento Kuwataka, Taiji Suzuki

机构 * Department of Mathematical Engineering(数学工程系) Information Physics, The University of Tokyo, Tokyo, Japan(信息物理,东京大学,东京,日本)

AI总结 本文提出测试时间训练与上下文学习结合的方法,通过理论分析证明该方法能有效适应不同任务中的特征向量和链接函数变化,提升模型预测性能。

Comments Under review at ICML 2026. 34 pages, 2 figures, appendix included; revised synthetic experiment and corrected mistakes

详情

展开后加载摘要…

URL PDF HTML 收藏
2506.03194 2026-02-03 cs.CV cs.AI cs.LG

HueManity: Probing Fine-Grained Visual Perception in MLLMs

HueManity: 探索 MLLMs 中的细粒度视觉感知

Rynaa Grover, Jayant Sravan Tamarapalli, Sahiti Yerramilli, Nilay Pande

机构 * Google(谷歌) Waymo

AI总结 HueManity 通过细粒度视觉感知基准测试揭示 MLLMs 在捕捉细粒度视觉细节方面的显著缺陷。

Journal ref ICML 2025 Workshop on Assessing World Models

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.04583 2026-02-03 cs.LG

Overcoming Spurious Solutions in Semi-Dual Neural Optimal Transport: A Smoothing Approach for Learning the Optimal Transport Plan

克服半对偶神经最优传输中的虚假解:一种平滑方法用于学习最优传输计划

Jaemoo Choi, Jaewoong Choi, Dohyun Kwon

机构 * Georgia Institute of Technology(佐治亚理工学院) Sungkyunkwan University(松均大学) University of Seoul(首尔大学) Korea Institute for Advanced Study(韩国高级研究院)

AI总结 OTP模型通过平滑方法克服半对偶神经OT中的虚假解问题,准确学习最优传输计划并提升图像到图像翻译性能。

Comments ICML 2025 (22 pages, 10 figures(

详情

展开后加载摘要…

URL PDF HTML 收藏