arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~

高校专区

ETH Zurich(苏黎世联邦理工学院)

2026-03-02 至 2026-03-02 共收录 5
2510.13328 2026-03-02 cs.LG cs.AI

Thompson Sampling via Fine-Tuning of LLMs

通过微调大语言模型进行汤普森采样

Nicolas Menet, Aleksandar Terzić, Michael Hersche, Andreas Krause, Abbas Rahimi

机构 * IBM Research – Zurich(IBM瑞士研究院) Department of Computer Science, ETH Zürich(苏黎世联邦理工学院计算机科学系)

AI总结 通过微调大语言模型实现高效汤普森采样,提升贝叶斯优化的样本和计算效率。

Comments accepted at ICLR 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2510.05228 2026-03-02 cs.LG cs.AI

CMT-Benchmark: A Benchmark for Condensed Matter Theory Built by Expert Researchers

CMT-Benchmark:由专家研究人员构建的凝聚态理论基准

Haining Pan, James V. Roggeveen, Erez Berg, Juan Carrasquilla, Debanjan Chowdhury, Surya Ganguli, Federico Ghimenti, Juraj Hasik, Henry Hunt, Hong-Chen Jiang, Mason Kamb, Ying-Jer Kao, Ehsan Khatami, Michael J. Lawler, Di Luo, Titus Neupert, Xiaoliang Qi, Michael P. Brenner, Eun-Ah Kim

机构 * Rutgers University(罗格斯大学) Harvard University(哈佛大学) Weizmann Institute of Science(魏茨曼科学研究所) ETH Zürich(苏黎世联邦理工学院) Cornell University(康奈尔大学) Stanford University(斯坦福大学) University of Zürich(苏黎世大学) Stanford Institute for Materials and Energy Sciences(斯坦福材料与能源科学研究所) SLAC National Accelerator Laboratory(斯坦福直线加速器实验室) University of California, Los Angeles(加州大学洛杉矶分校) National Taiwan University(台湾大学) San José State University(圣何塞州立大学)

AI总结 CMT-Benchmark通过专家设计的50个凝聚态理论问题评估LLM在物理推理能力上的不足,揭示当前模型在复杂科学问题上的局限性。

Comments CMT-Benchmark dataset is available at https://huggingface.co/datasets/JVRoggeveen/cmt_benchmark. CMT-Benchmark was referenced in the Gemini 3 Deep Think (February 2026) release at https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-deep-think/

Journal ref International Conference on Learning Representations (ICLR) main conference 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.18679 2026-03-02 cs.CV

Efficient Degradation-agnostic Image Restoration via Channel-Wise Functional Decomposition and Manifold Regularization

通过通道级功能分解和流形正则化实现高效的退化无关图像恢复

Bin Ren, Yawei Li, Xu Zheng, Yuqian Fu, Danda Pani Paudel, Hong Liu, Ming-Hsuan Yang, Luc Van Gool, Nicu Sebe

机构 * Mohamed bin Zayed University of Artificial Intelligence(莫扎德·本·泽德人工智能大学) University of Trento(特伦托大学) ETH Zürich(苏黎世联邦理工学院) HKUST (GZ)(香港科技大学(广州)) Peking University(北京大学) University of California, Merced(加州大学默塞德分校)

AI总结 MIRAGE通过通道级功能分解和流形正则化,在高效性与性能之间取得平衡,实现退化无关图像恢复的先进性能。

Comments Accepted by ICLR'2026, All-in-One Image Restoration, low-level vision, Transformer

详情

展开后加载摘要…

URL PDF HTML 收藏
2306.09778 2026-03-02 cs.LG cs.NA math.NA math.OC stat.ML

Gradient is All You Need? How Consensus-Based Optimization can be Interpreted as a Stochastic Relaxation of Gradient Descent

梯度是全部需要吗?如何将基于共识的优化解释为梯度下降的随机松弛

Konstantin Riedl, Timo Klock, Carina Geldhauser, Massimo Fornasier

机构 * University of Oxford, Mathematical Institute(牛津大学数学研究所) Deeptech Consulting(德普科技咨询) ETH Zurich, Department of Mathematics(苏黎世联邦理工学院数学系) Technical University of Munich, School of Computation, Information and Technology, Department of Mathematics(慕尼黑技术大学计算、信息与技术学院数学系) Munich Center for Machine Learning Munich Data Science Institute(慕尼黑机器学习中心慕尼黑数据科学研究所)

AI总结 本文将基于共识的优化解释为梯度下降的随机松弛,揭示其在非凸优化中的全局收敛性及对能量壁垒的克服能力。

Comments 49 pages, 5 figures

详情

展开后加载摘要…

URL PDF HTML 收藏
2208.14960 2026-03-02 stat.ME cs.LG math.ST stat.ML stat.TH

Stationary Kernels and Gaussian Processes on Lie Groups and their Homogeneous Spaces I: the compact case

平稳核与李群及其齐性空间上的高斯过程 I:紧致情况

Iskander Azangulov, Andrei Smolensky, Alexander Terenin, Viacheslav Borovitskiy

机构 * St. Petersburg State University and University of Oxford(圣彼得堡国立大学和牛津大学) St. Petersburg State University and Neapolis University Pafos(圣彼得堡国立大学和纳皮奥斯大学帕福斯) University of Cambridge and Cornell University(剑桥大学和康奈尔大学) ETH Zürich(苏黎世联邦理工学院)

AI总结 本文研究了在李群及其齐性空间上构建平稳高斯过程的技术,特别针对紧致空间,提供计算方法使其与现有高斯过程软件兼容。

Comments This version fixes two mathematical typos, in equations (58) and (65), where both sums should be taken only over the diagonal part $π^{(λ)}_{jj}$ and not over $π^{(λ)}_{jk}$ as had erroneously been written in the previous version. The proofs for both statements remain unchanged. We thank Nathaël Da Costa for making us aware of this pair of typos

Journal ref Journal of Machine Learning Research, 2024

详情

展开后加载摘要…

URL PDF HTML 收藏