Reproduce “RAGCompass: Generalizable Retriever Routing via Reranking” 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
Valentin Six, Mufei Li, Dongqi Fu, Yinglong Xia, Hong Li, Pan Li
RAGCompass routes among retrieved passage sets by scoring how well each supports an answer. A shared model reads each query-passage set, and training uses answer F1 as utility. On queries with a top-two retriever answer-score gap of at least 0.05, it scores 0.3552 versus 0.2782 for the best fixed retriever; across all queries, the gain is 2.28 percentage points.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.