List your company

Question Answering on Book Corpus Multi-hop Private (176 questions)

89.8AccuracyNWM Graph Retrieval
-3.59220.65444.969.146Jul 6, 2026

Evaluation results

MethodLinks
NWM Graph RetrievalAnswerer model=Opus 4.8, Extractor model=Sonnet 4.52026.0789.8
GraphitiAnswerer model=Opus 4.8, Extractor model=Sonnet 4.52026.0757.4
NWM State MemoryAnswerer model=Opus 4.8, Extractor model=Sonnet 4.52026.0735.8
GraphRAGAnswerer model=Opus 4.82026.0718.8
RAG (BGE+BM25+RRF)Answerer model=Opus 4.82026.0717.6
Simple SearchAnswerer model=Opus 4.82026.078.5
No MemoryAnswerer model=Opus 4.82026.070