Reproduce “Re$^3$Rank: Listwise Reranking with Reasoning-Conditioned Retrieval Heads” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ruiyi Yan, Zhuoyuan Mao, Yugo Murawaki, Yiwen Guo
Re³Rank reranks retrieved passages by using a generated trace, formed after reading the question and candidate passages, to guide attention-based scoring. In a MuSiQue example, the trace links a painting’s creator to a separate birthday passage and ranks it first. Across nine datasets, R@3 reaches 76.70 over QRRanker’s 73.39 on the top-50 pool from one retriever; other pools remain untested.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.