作者位次 #6/7SIGIR 2026Oral· Information Retrieval (Full)
ARHN: Answer-Centric Relabeling of Hard Negatives with Open-Source LLMs for Dense Retrieval
Hyewon Choi · Jooyoung Choi · Hansol Jang · Hyun Kim · Chulmin Yun · Changwook Jun · Stanley Jungkyu Choi
摘要
Neural retrievers are often trained on large-scale triplet data comprising a query, a positive passage, and a set of hard negatives. In practice, hard-negative mining can introduce false negatives and other ambiguous negatives, including passages that are relevant or contain partial answers to th...