arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

University of Texas at Austin(得克萨斯大学奥斯汀分校)

2026-05-01 至 2026-05-01 共收录 3
2310.02277 2026-05-01 cs.LG cs.AI

Junk DNA Hypothesis: Pruning Small Pre-Trained Weights Irreversibly and Monotonically Impairs "Difficult" Downstream Tasks in LLMs

垃圾DNA假说:修剪小的预训练权重不可逆且单调地损害LLM中的“困难”下游任务

Lu Yin, Ajay Jaiswal, Shiwei Liu, Souvik Kundu, Zhangyang Wang

机构 * University of Surrey Eindhoven University of Technology University of Texas at Austin Intel Labs University of Oxford

AI总结 该研究提出垃圾DNA假说,指出LLM预训练权重中存在关键知识,修剪小权重会单调损害困难下游任务性能,且即使允许持续训练也无法弥补损失。

Comments Published at ICML 2024

详情

展开后加载摘要…

URL PDF HTML 收藏
2603.06526 2026-05-01 cond-mat.mtrl-sci cs.LG

Predicting Atomistic Transitions with Transformers

用变压器预测原子级转变

Henry Tischler, Wenting Li, Qi Tang, Danny Perez, Thomas Vogel

机构 * Computing and Artificial Intelligence Division, Los Alamos National Laboratory(计算与人工智能部门,洛斯阿拉莫斯国家实验室) School of Engineering and Computer Science, University of Denver(工程与计算机科学学院,丹佛大学) Department of Physics and Astronomy, University of Denver(物理与天文学系,丹佛大学) Department of Electrical and Computer Engineering, University of Texas at Austin(电气与计算机工程系,德克萨斯大学奥斯汀分校) School of Computational Science and Engineering, Georgia Institute of Technology(计算科学与工程学院,佐治亚理工学院) Theoretical Division, Los Alamos National Laboratory(理论部门,洛斯阿拉莫斯国家实验室) X Computational Physics Division, Los Alamos National Laboratory(X计算物理部门,洛斯阿拉莫斯国家实验室)

AI总结 本文利用变压器模型高效预测纳米簇中的原子级转变,通过评估物理有效性并生成多种微态,降低计算成本。

Comments Presented at the 2025 Conference on Data Analysis (CoDA), February 25-28, Santa Fe, New Mexico

详情

展开后加载摘要…

URL PDF HTML 收藏
2411.18796 2026-05-01 cs.LG q-bio.QM

Graph-Based Biomarker Discovery and Interpretation for Alzheimer's Disease

基于图的阿尔茨海默病生物标志物发现与解释

Maryam Khalid, Fadeel Sher Khan, John Broussard, Arko Barman

机构 * Rice University(里士大学) University of Texas at Austin(德克萨斯大学奥斯汀分校) University of Texas Health Science Center at Houston(德克萨斯大学健康科学中心休斯顿分校)

AI总结 本文提出BRAIN框架,通过图表示方法联合优化诊断准确性和生物标志物发现,揭示阿尔茨海默病诊断相关的生物标志物子网络。

详情

展开后加载摘要…

URL PDF HTML 收藏