arXivDaily arXiv每日学术速递 周一至周五更新

高校专区

Harvard University(哈佛大学)

2026-06-05 至 2026-06-05 共收录 5
2606.05436 2026-06-05 cs.AI cs.CL cs.IR

Ten Headache Specialists versus Artificial Intelligence for Clinical Literature Summarization: A Critical Evaluation and Comparison

十位头痛专家与人工智能在临床文献总结中的比较:一项关键评估与对比

Alejandro Lozano, Keiko Ihara, Ping-Hao Yang, Carrie E. Robertson, Jennifer Stern, Allan Purdy, Hsiangkuo Yuan, Pengfei Zhang, Yulia Orlova, Olga Fermo, Jennifer Hranilovich, Fred Cohen, Todd J. Schwedt, Jenelle A. Jindal, Serena Yeung-Levy, Chia-Chun Chiang

机构 * Stanford University Palo Alto CA USA(斯坦福大学) Department of Neurology Mayo Clinic Rochester MN USA(梅奥诊所神经科) Department of Neurology Dalhousie University Halifax Canada(达尔豪斯大学神经科) Jefferson Headache Center Department of Neurology Thomas Jefferson University PA USA(泰勒大学神经科) Beth Israel Deaconess Medical Center Boston MA USA(贝斯以色列医疗中心) Department of Neurology University of Florida Gainesville FL USA(佛罗里达大学神经科) University of Colorado School of Medicine Department of Pediatrics Division of Child Neurology Aurora CO USA(科罗拉多医学院儿科部儿童神经科) Department of Medicine Mount Sinai Hospital Icahn School of Medicine at Mount Sinai New York NY USA(西奈医院医学部) Department of Neurology Mayo Clinic Scottsdale AZ USA(梅奥诊所Scottsdale分部) Harvard Medical School Boston MA USA(哈佛医学院) Department of Neurology Mount Sinai Hospital Icahn School of Medicine at Mount Sinai New York NY USA(西奈医院神经科)

AI总结 本研究通过构建基于RAG的AI框架,比较了三种大语言模型与十位头痛专家在临床文献总结方面的表现,发现专家撰写的摘要更受青睐,但专家有时难以区分人类与AI生成的摘要。

详情

展开后加载摘要…

URL PDF HTML 收藏
2605.04135 2026-06-05 cs.CY cs.AI cs.CL

Frontier Lag: A Bibliometric Audit of Capability Misrepresentation in Academic AI Evaluation

前沿滞后:学术AI评估中能力误述的文献计量审计

David Gringras, Misha Salahshoor

机构 * Harvard University(哈佛大学) AISST

AI总结 通过审计112,303篇LLM相关论文,发现中位论文评估的模型落后同期前沿10.85 ECI(约1.4倍Claude Sonnet 3.7与Claude Opus 4.5的差距),且差距以每年5.53 ECI扩大,仅3.2%的摘要披露推理模式状态,52.5%的结论将结果泛化为“AI”,并提出VERSIO-AI检查表等补救措施。

Comments v2. 65 pp, 9 figs, 8 tables, 8 appendices. Pre-registered on OSF: doi.org/10.17605/OSF.IO/7XM3D. Code+data: doi.org/10.5281/zenodo.20060457. VERSIO-AI v1.2 reporting checklist (Appendix A): doi.org/10.5281/zenodo.20060459. frontierlag package + per-DOI audit tool: frontierlag.org

详情

展开后加载摘要…

URL PDF HTML 收藏
2602.10314 2026-06-05 cs.LG

Stop Training for the Worst: Progressive Unmasking Accelerates Masked Diffusion Training

停止训练于最差:渐进性解蔽加速了掩码扩散训练

Jaeyeon Kim, Jonathan Geuter, David Alvarez-Melis, Sham Kakade, Sitan Chen

机构 * Harvard University(哈佛大学) Kempner Institute(凯普纳研究所)

AI总结 本文提出了一种名为渐进性解蔽(PUMA)的方法,通过修改前向掩码过程,使训练时间和推理时的掩码模式一致,从而加速了掩码扩散模型的训练。

详情

展开后加载摘要…

URL PDF HTML 收藏
2508.04409 2026-06-05 stat.ML cs.LG

The Relative Instability of Model Comparison with Cross-validation

模型比较与交叉验证的相对不稳定性

Alexandre Bayle, Lucas Janson, Lester Mackey

机构 * Department of Statistics, Harvard University, Cambridge, MA, USA(哈佛大学统计系) Microsoft Research New England, Cambridge, MA, USA(微软研究院新英格兰分部)

AI总结 研究指出即使个体稳定的模型在比较时也可能产生相对不稳定的结果,挑战了交叉验证推断的有效性,特别指出Lasso和软阈值化在最有利的学习条件下仍会导致无效的交叉验证推断。

详情

展开后加载摘要…

URL PDF HTML 收藏
2305.12640 2026-06-05 cs.AI cs.LG stat.ML

Limited Resource Allocation in a Non-Markovian World: The Case of Maternal and Child Healthcare

在非马尔可夫世界中的有限资源分配:产科与儿童保健的案例

Panayiotis Danassis, Shresth Verma, Jackson A. Killian, Aparna Taneja, Milind Tambe

机构 * Harvard University(哈佛大学) Google Research(谷歌研究)

AI总结 本文研究了在非马尔可夫环境下如何通过时间序列方法优化资源分配,提出了一种新的时间序列臂排名指数(TARI)策略,以提高产科和儿童保健项目的参与度和依从性。

Comments Proceedings of the 32nd International Joint Conference on Artificial Intelligence (IJCAI 2023)

详情

展开后加载摘要…

URL PDF HTML 收藏