arXivDaily arXiv每日学术速递 周一至周五更新

期刊&会议

Conference on Empirical Methods in Natural Language Processing · 会议 · Natural Language Processing

2026-02-02 至 2026-02-02 共收录 3
2509.16648 2026-02-02 cs.AI cs.CL cs.LG

FESTA: Functionally Equivalent Sampling for Trust Assessment of Multimodal LLMs

FESTA:用于多模态大语言模型信任评估的功能等效采样

Debarpan Bhattacharya, Apoorva Kulkarni, Sriram Ganapathy

机构 * Indian Institute of Science(印度科学研究院) University of Maryland College Park(马里兰大学学院公园分校)

AI总结 FESTA通过功能等效采样技术提升多模态大语言模型的预测选择性性能,实现33.3%和29.6%的改进。

Comments Accepted in the Findings of EMNLP, 2025

Journal ref EMNLP 2025

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.16245 2026-02-02 cs.CL

Diverse, not Short: A Length-Controlled Data Selection Strategy for Improving Response Diversity of Language Models

多样而非简短:一种长度控制的数据选择策略以提高语言模型的响应多样性

Vijeta Deshpande, Debasmita Ghose, John D. Patterson, Roger Beaty, Anna Rumshisky

机构 * University of Massachusetts Lowell(马萨诸塞大学洛维尔分校) Yale University(耶鲁大学) Pennsylvania State University(宾夕法尼亚州立大学) Amazon AGI(亚马逊人工智能研究院)

AI总结 Diverse-NS通过长度控制的数据选择策略提升语言模型响应多样性,适用于创造性生成任务,并在不同规模模型间展示出显著效果。

Comments Accepted to EMNLP 2025 Main

详情

展开后加载摘要…

URL PDF HTML 收藏
2505.13718 2026-02-02 cs.AI cs.CL

Warm Up Before You Train: Unlocking General Reasoning in Resource-Constrained Settings

在训练前热身:在资源受限环境下解锁通用推理

Safal Shrestha, Minwu Kim, Aadim Nepal, Anubhav Shrestha, Keith Ross

机构 * Department of Computer Science, New York University Abu Dhabi(纽约大学阿布扎克分校计算机科学系)

AI总结 本文提出一种分两阶段的训练策略,在资源受限环境下通过热身提升大语言模型的推理能力。

Comments Accepted to EMNLP 2025

Journal ref Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing (EMNLP 2025)

详情

展开后加载摘要…

URL PDF HTML 收藏