arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

University of Washington(华盛顿大学)

2026-07-23 至 2026-07-23 共收录 5
2607.20251 2026-07-23 cs.CL 新提交

Exposure is Optional: Learning Unlike Coordination in Language Models

曝光非必需:语言模型中学习不同类别的并列结构

Jiamu Luo, Shane Steinert-Threlkeld

机构 * University of Washington(华盛顿大学)

AI总结 研究语言模型中不同类并列结构习得是否需直接曝光,用过滤语料库训练GPT-2模型,发现无需直接曝光,模型能泛化处理,还揭示了模型处理方式及可从同类并列结构学习,助力理解语言模型结构表示。

Comments 13 pages, 6 tables, 2 figures, to submit to TACL

详情

展开后加载摘要…

URL PDF HTML 收藏
2607.19854 2026-07-23 cs.LG stat.ML 新提交

Asymptotically Optimal Regret for Reinforcement Learning without Horizon Dependence

无时间范围依赖的强化学习的渐近最优遗憾值

Runlong Zhou, Zihan Zhang, Maryam Fazel, Simon S. Du

机构 * University of Washington(华盛顿大学) Hong Kong University of Science and Technology(香港科技大学)

AI总结 研究有限时间范围齐次表格马尔可夫决策过程的无时间范围遗憾值最小化,提出新算法,证明遗憾值上界\( \tilde O(\sqrt{SAK}+S^8A^3) \),渐近最优且消除\( \log H \)依赖,改进先前结果。

Comments 78 pages

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.03825 2026-07-23 cs.LG 版本更新

The Geometry of Learning to Avoid Interventions

从紧急停止干预中学习鲁棒性干预

Ethan Pronovost, Khimya Khetarpal, Siddhartha Srinivasa

机构 * Paul G. Allen School of Computer Science \& Engineering, University of Washington, Seattle, USA Google DeepMind, Seattle, USA

AI总结 本文提出残差干预微调算法,通过结合先验策略解决干预信号不明确的问题,实现鲁棒的策略改进。

详情

展开后加载摘要…

URL PDF HTML 收藏
2211.09949 2026-07-23 cs.CL cs.LG cs.SD eess.AS

Is Smaller Always Faster? Tradeoffs in Compressing Self-Supervised Speech Transformers

更小的总是更快的吗?压缩自监督语音Transformer的权衡

Tzu-Quan Lin, Tsung-Huan Yang, Chun-Yao Chang, Kuang-Ming Chen, Tzu-hsun Feng, Hung-yi Lee, Hao Tang

机构 * Graduate Institute of Communication Engineering, National Taiwan University, Taiwan(台湾国立台湾大学通信工程研究所) University of California, Los Angeles, United States(美国加州大学洛杉矶分校) University of Washington, United States(美国华盛顿大学) University of Edinburgh, United Kingdom(英国爱丁堡大学)

AI总结 本文研究了压缩自监督语音Transformer的权衡,评估了四种压缩方法并比较了几种最新技术,为实际部署提供指导。

Comments Accepted at ASRU 2025. Code is available at https://github.com/nervjack2/Speech-SSL-Compression

Journal ref 2025 IEEE Automatic Speech Recognition and Understanding Workshop (ASRU), pp. 1-7

详情

展开后加载摘要…

URL PDF HTML 收藏
2502.10930 2026-07-23 cs.LG math.DS

Reduced Order Modeling with Shallow Recurrent Decoder Networks

基于浅层递归解码器网络的降阶建模

Matteo Tomasetto, Jan P. Williams, Francesco Braghin, Andrea Manzoni, J. Nathan Kutz

机构 * Department of Mechanical Engineering, Politecnico di Milano, Milano, Italy(米兰理工学院机械工程系) Department of Mechanical Engineering, University of Washington, Seattle, WA(华盛顿大学机械工程系) MOX - Department of Mathematics, Politecnico di Milano, Milano, Italy(米兰理工学院数学系) Department of Applied Mathematics and Electrical and Computer Engineering, University of Washington, Seattle, WA(华盛顿大学应用数学与电气与计算机工程系)

AI总结 本文提出SHRED-ROM,一种基于浅层递归解码器网络的降阶建模方法,能够高效重建高维状态动态,处理多种参数依赖,并准确估计未知参数。

详情

展开后加载摘要…

URL PDF HTML 收藏