Abstention on HotpotQA
76.8Abstain AccuracyR2C
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| R2CRAG Framework=ReAct, Confidence Threshold=0.92025.10 | 76.8 | 85.08 | |
| SelfCRAG Framework=ReAct, Confidence Threshold=0.92025.10 | 75.8 | 83.8 | |
| ReaCRAG Framework=ReAct, Confidence Threshold=0.92025.10 | 75.4 | 83.17 | |
| R2CRAG Framework=ReSearch, Confidence Threshold=0.92025.10 | 74.8 | 78.71 | |
| R2CRAG Framework=SelfAsk, Confidence Threshold=0.92025.10 | 74.4 | 81.76 | |
| R2CRAG Framework=Search-o1, Confidence Threshold=0.92025.10 | 74.4 | 82.22 | |
| R2CRAG Framework=Search-R1, Confidence Threshold=0.92025.10 | 73.4 | 77.57 | |
| RrrCRAG Framework=ReAct, Confidence Threshold=0.92025.10 | 72.2 | 80.82 | |
| ReaCRAG Framework=Search-o1, Confidence Threshold=0.92025.10 | 71.4 | 79.48 | |
| P(true)RAG Framework=ReSearch, Confidence Threshold=0.92025.10 | 71.4 | 73.46 | |
| P(true)RAG Framework=SelfAsk, Confidence Threshold=0.92025.10 | 70.8 | 75.33 | |
| P(true)RAG Framework=Search-R1, Confidence Threshold=0.92025.10 | 70 | 71.15 | |
| SelfCRAG Framework=Search-o1, Confidence Threshold=0.92025.10 | 69 | 78.07 | |
| ReaCRAG Framework=Search-R1, Confidence Threshold=0.92025.10 | 68.8 | 71.11 | |
| SelfCRAG Framework=ReSearch, Confidence Threshold=0.92025.10 | 68.6 | 75.19 | |
| ReaCRAG Framework=ReSearch, Confidence Threshold=0.92025.10 | 68.6 | 72.6 | |
| P(true)RAG Framework=ReAct, Confidence Threshold=0.92025.10 | 67.2 | 73.46 | |
| RrrCRAG Framework=Search-o1, Confidence Threshold=0.92025.10 | 66.2 | 71.5 | |
| P(true)RAG Framework=Search-o1, Confidence Threshold=0.92025.10 | 65.8 | 71.73 | |
| RrrCRAG Framework=SelfAsk, Confidence Threshold=0.92025.10 | 65.6 | 70.54 | |
| ReaCRAG Framework=SelfAsk, Confidence Threshold=0.92025.10 | 64.6 | 70.45 | |
| RrrCRAG Framework=ReSearch, Confidence Threshold=0.92025.10 | 64.4 | 63.37 | |
| SelfCRAG Framework=SelfAsk, Confidence Threshold=0.92025.10 | 64.2 | 69.51 | |
| RrrCRAG Framework=Search-R1, Confidence Threshold=0.92025.10 | 64 | 61.2 | |
| SelfCRAG Framework=Search-R1, Confidence Threshold=0.92025.10 | 62.8 | 62.34 |