Reproduce “Beyond Retrieval Failure: Understanding Why LLMs Miss Evidence in Multi-Hop RAG” 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
Jiaqi Wei, Wenxuan Huang, Chengdong Xu, Xiang Zhang, Wanli Ouyang
ExecutionRAG addresses search drift in multi-hop retrieval-augmented generation by tracking unanswered sub-questions and linking each to evidence. Unlike changing the retriever, its register guides which requirement the agent searches for next. In 1,000 MuSiQue questions, missing-evidence failures fell from 69.3% of A-RAG failures to 43.9% of ExecutionRAG failures with MiniLM; this aggregate comparison does not measure search-drift prevalence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.