Question Answering on HotpotQA (EM, F1, EFR)
58.4EMPruneRAG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| PruneRAGBackbone=Qwen3-32B, Time=22212026.01 | 58.4 | 66 | 24.6 | |
| PruneRAGBackbone=Llama-3.1-70B-Instruct, Time=14322026.01 | 58.2 | 66 | 29.2 | |
| RAGBackbone=Qwen3-32B, Time=13942026.01 | 54.2 | 62.6 | 21.9 | |
| RAGBackbone=Llama-3.1-70B-Instruct, Time=6032026.01 | 50.6 | 57.2 | 29.7 | |
| ConTRGenBackbone=Llama-3.1-70B-Instruct, Time=227992026.01 | 50.6 | 57.7 | 30.7 | |
| ConTRGenBackbone=Qwen3-32B, Time=169212026.01 | 49.8 | 58.7 | 36.4 | |
| Search-o1Backbone=Llama-3.1-70B-Instruct, Time=30572026.01 | 46.2 | 52.6 | 40.2 | |
| Search-o1Backbone=Qwen3-32B, Time=50472026.01 | 42.4 | 49.4 | 43.6 | |
| RAG-StarBackbone=Llama-3.1-70B-Instruct, Time=432392026.01 | 41.6 | 50.9 | 49.6 | |
| ReactBackbone=Llama-3.1-70B-Instruct, Time=28242026.01 | 41.2 | 46.6 | 41.8 | |
| RAG-StarBackbone=Qwen3-32B, Time=525272026.01 | 38.4 | 46.9 | 49.3 | |
| ProbTreeBackbone=Llama-3.1-70B-Instruct, Time=25832026.01 | 36.4 | 44.6 | 50.5 | |
| ReactBackbone=Qwen3-32B, Time=79272026.01 | 34.4 | 40.4 | 35.5 | |
| Self-RAGBackbone=Llama-3.1-70B-Instruct, Time=52942026.01 | 30.6 | 37.5 | 59.1 | |
| Self-RAGBackbone=Qwen3-32B, Time=52942026.01 | 30.6 | 37.5 | 59.1 | |
| ProbTreeBackbone=Qwen3-32B, Time=25042026.01 | 25.6 | 38.1 | 59 | |
| VanillaBackbone=Llama-3.1-70B-Instruct, Time=5152026.01 | 25 | 30.8 | — | |
| VanillaBackbone=Qwen3-32B, Time=34592026.01 | 21.2 | 25.4 | — | |
| MemoRAGBackbone=Llama-3.1-70B-Instruct, Time=10002026.01 | 10 | 10 | 10 | |
| MemoRAGBackbone=Qwen3-32B, Time=10002026.01 | 10 | 10 | 10 |