List your company
ResearchFormerly WizWand

Long-form Question Answering on Novel GraphRAG-Bench

85.3LLM-AccA-RAG (Full)
26.5441.79557.0572.305Feb 3, 2026
Updated 8mo ago

Evaluation results

MethodLinks
A-RAG (Full)Backbone LLM=GPT-5-mini2026.0285.3
A-RAG (Naive)Backbone LLM=GPT-5-mini2026.0280.4
GraphRAGBackbone LLM=GPT-5-mini2026.0277.1
A-RAG (Full)Backbone LLM=GPT-4o-mini2026.0272.7
Naive RAGBackbone LLM=GPT-5-mini2026.0270.6
HippoRAG2Backbone LLM=GPT-4o-mini2026.0270.1
A-RAG (Naive)Backbone LLM=GPT-4o-mini2026.0270
Naive RAGBackbone LLM=GPT-4o-mini2026.0268.5
RAGentABackbone LLM=GPT-4o-mini2026.0261.3
FaithfulRAGBackbone LLM=GPT-5-mini2026.0260.7
RAGentABackbone LLM=GPT-5-mini2026.0260.2
LinearRAGBackbone LLM=GPT-5-mini2026.0254.7
HippoRAG2Backbone LLM=GPT-5-mini2026.0254.3
LinearRAGBackbone LLM=GPT-4o-mini2026.0245.4
Direct AnswerBackbone LLM=GPT-4o-mini2026.0245.3
Direct AnswerBackbone LLM=GPT-5-mini2026.0245.1
MA-RAGBackbone LLM=GPT-5-mini2026.0245.1
MA-RAGBackbone LLM=GPT-4o-mini2026.0244.5
FaithfulRAGBackbone LLM=GPT-4o-mini2026.0233.3
GraphRAGBackbone LLM=GPT-4o-mini2026.0228.8