Reproduce “GRASP: Query-Adaptive Graph Retrieval framework with Cross-Attention Scoring and Path Sufficiency Verification for Question Answering” 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

Rui Xu, Le Wang, Tianyue Wang

GRASP reduces unsupported claim rates by checking whether retrieved paths contain enough evidence before generation. It scores edges, stops adaptively, and checks path sufficiency, retrying retrieval when needed. Across three biomedical benchmarks, it lowers unsupported claim rate by 8.9–13.2 percentage points versus MINERVA; on MedKGent-KG, the rate falls from 24.0% to 15.1%. This measures path support, not factual error or clinical safety.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.