arXivDaily arXiv每日学术速递 周一至周五更新

期刊&会议

ACM SIGIR Conference on Research and Development in Information Retrieval · 会议 · Information Retrieval

2026-05-26 至 2026-05-26 共收录 4
2509.16931 2026-05-26 cs.IR cs.AI cs.LG

Equip Pre-ranking with Target Attention by Residual Quantization

通过残差量化为预排序阶段配备目标注意力机制

Yutong Li, Yu Zhu, Yichen Qiao, Ziyu Guan, Lv Shao, Tong Liu, Bo Zheng

机构 * Taobao \& Tmall Group of Alibaba Hangzhou China Shanghai Jiao Tong University Shanghai China Xidian University Xi'an China Taobao \& Tmall Group of Alibaba Beijing China Taobao \& Tmall Group of Alibaba Shanghai Jiao Tong University Xidian University

AI总结 提出TARQ框架,利用残差量化在预排序阶段近似目标注意力架构,首次在延迟关键阶段引入TA建模能力,实现精度与效率的新最优平衡。

Comments 5 pages, 2 figures, accepted by SIGIR 2026 Short Paper Track

详情

展开后加载摘要…

URL PDF HTML 收藏
2604.23396 2026-05-26 cs.IR cs.AI cs.CL cs.LG

Lost in Decoding? Reproducing and Stress-Testing the Look-Ahead Prior in Generative Retrieval

迷失在解码中?复现与压力测试生成式检索中的前瞻先验

Kidist Amde Mekonnen, Yongkang Li, Yubao Tang, Simon Lupart, Maarten de Rijke

机构 * University of Amsterdam(阿姆斯特丹大学)

AI总结 本文复现并压力测试了生成式检索中的前瞻先验方法PAG,发现其规划信号在词汇表面形式变化下脆弱,并评估了跨语言鲁棒性与查询端缓解策略。

Comments 12 pages, 5 figures, 9 tables; accepted to the 49th International ACM SIGIR Conference on Research and Development in Information Retrieval, July 20-24, 2026, Melbourne/Naarm, Australia

Journal ref Proceedings of the 49th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR '26), pages XXX-XXX, 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2601.22925 2026-05-26 cs.IR cs.AI cs.LG

BEAR: Towards Beam-Search-Aware Optimization for Recommendation with Large Language Models

BEAR:面向大语言模型推荐中束搜索感知的优化

Weiqin Yang, Bohao Wang, Zhenxiang Xu, Jiawei Chen, Shengjia Zhang, Jingbang Chen, Canghong Jin, Can Wang

机构 * Zhejiang University(浙江大学) The Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳)) Hangzhou City University(杭州市城市大学)

AI总结 针对监督微调与束搜索推理之间的不一致性,提出BEAR正则化方法,通过确保正例每个token在解码步骤中排名前B来避免过早剪枝,显著提升推荐性能。

Comments Accepted by SIGIR 2026

详情

展开后加载摘要…

URL PDF HTML 收藏
2504.05181 2026-05-26 cs.IR cs.AI cs.DL cs.LG

Lightweight and Direct Document Relevance Optimization for Generative Information Retrieval

轻量级直接文档相关性优化用于生成式信息检索

Kidist Amde Mekonnen, Yubao Tang, Maarten de Rijke

机构 * Institute for Clarity in Documentation(文档清晰度研究所) Inria Paris-Rocquencourt(巴黎- Rocquencourt 国家信息与自动化所) Rajiv Gandhi University(拉朱·甘地大学) Tsinghua University(清华大学) Palmer Research Laboratories(帕勒尔研究实验室) University of Amsterdam(阿姆斯特丹大学)

AI总结 提出直接文档相关性优化(DDRO)方法,通过成对排序直接对齐令牌级文档ID生成与文档级相关性估计,无需显式奖励建模和强化学习,在MS MARCO和Natural Questions上分别提升MRR@10 7.4%和19.9%。

Comments 12 pages, 3 figures. SIGIR '25 Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval July 13--18, 2025 Padua, Italy. Code and pretrained models available at: https://github.com/kidist-amde/ddro/

Journal ref Proceedings of the 48th International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR '25), pages 1327-1338, 2025

详情

展开后加载摘要…

URL PDF HTML 收藏