TimelyRAG：面向重叠演化文档的语义-时间混合检索
原标题：TimelyRAG: Semantic-Temporal Hybrid Retrieval for Time-Critical Question Answering in Overlapping-Evolving Documents
AI 摘要
KAIST 的研究者提出 TimelyRAG，一个与检索器无关的语义-时间混合重排序框架，将时间距离纳入排序，以应对法律、政策等文档通过修正案重叠演化的场景。同时发布 TimelyQABench，首个针对重叠演化监管语料的时间敏感问答基准。实验显示 nDCG@10 最高提升 28.6%，Hit@10 提升 19.1%。
正文节选
TimelyRAG: Semantic-Temporal Hybrid Retrieval for Time-Critical Question Answering in Overlapping-Evolving Documents Abstract Although large language models (LLMs) and retrieval-augmented generation (RAG) have advanced open-domain question answering (QA), they remain unreliable when documents evolve through amendments. Existing time-sensitive retrieval methods address only the disjoint-evolving environment, where each update is an independent snapshot. However, laws, policies, and regulations of