RAGPPI: RAG Benchmark for Protein-Protein Interactions in Drug Discovery
Researchers have introduced RAGPPI, a new benchmark designed to evaluate Retrieval-Augmented Generation (RAG) systems for identifying the biological impacts of protein-protein interactions (PPIs) in drug discovery. The benchmark consists of 4,420 question-answer pairs, with a gold-standard subset of 500 pairs created through expert annotation and a silver-standard set generated using an ensemble auto-evaluation LLM. RAGPPI aims to advance RAG systems for drug discovery applications by providing a dedicated resource for this specific task. AI