Question Answering on HotpotQA 1,000 questions sample (val)
66.3Containment AccuracyFlowRAG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| FlowRAGLLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v22026.06 | 66.3 | 68.6 | |
| LinearRAGLLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v22026.06 | 65.5 | 66.8 | |
| HippoRAG2LLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v22026.06 | 62.9 | 64.3 | |
| GFM-RAGLLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v22026.06 | 62.7 | 65.6 | |
| KGPLLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v22026.06 | 61.5 | 60.9 | |
| E2GraphRAGLLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v22026.06 | 61 | 63.9 | |
| LightRAGLLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v22026.06 | 60.3 | 59.5 | |
| HippoRAGLLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v22026.06 | 57 | 59.3 | |
| RAPTORLLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v22026.06 | 55.9 | 58.3 | |
| Vanilla RAGLLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v2, Prompting=CoT2026.06 | 55.7 | 58.6 | |
| G-retrieverLLM=GPT-4o-mini, k=5, Embedding model=all-mpnet-base-v22026.06 | 42.2 | 40.6 |