List your company
ResearchFormerly WizWand

Long-form Question Answering on GraphRAG-Bench Med

93.1LLM AccuracyA-RAG (Full)
40.47654.13867.881.462Feb 3, 2026
Updated 8mo ago

Evaluation results

MethodLinks
A-RAG (Full)Backbone LLM=GPT-5-mini2026.0293.1
A-RAG (Naive)Backbone LLM=GPT-5-mini2026.0292.7
Direct AnswerBackbone LLM=GPT-5-mini2026.0290.5
GraphRAGBackbone LLM=GPT-5-mini2026.0287.3
Naive RAGBackbone LLM=GPT-5-mini2026.0286.1
A-RAG (Full)Backbone LLM=GPT-4o-mini2026.0279.4
LinearRAGBackbone LLM=GPT-5-mini2026.0279.2
A-RAG (Naive)Backbone LLM=GPT-4o-mini2026.0279
HippoRAG2Backbone LLM=GPT-5-mini2026.0278.2
FaithfulRAGBackbone LLM=GPT-5-mini2026.0275.4
Naive RAGBackbone LLM=GPT-4o-mini2026.0275.3
RAGentABackbone LLM=GPT-5-mini2026.0273.7
HippoRAG2Backbone LLM=GPT-4o-mini2026.0272
Direct AnswerBackbone LLM=GPT-4o-mini2026.0268.6
MA-RAGBackbone LLM=GPT-5-mini2026.0268.3
RAGentABackbone LLM=GPT-4o-mini2026.0267.7
MA-RAGBackbone LLM=GPT-4o-mini2026.0262.3
LinearRAGBackbone LLM=GPT-4o-mini2026.0253.1
GraphRAGBackbone LLM=GPT-4o-mini2026.0251.3
FaithfulRAGBackbone LLM=GPT-4o-mini2026.0242.5