Question Answering on TriviaQA (Exact Match Accuracy)
75.4EM AccuracyICRL
Evaluation results
| Method | Links | |
|---|---|---|
| ICRLModel=Qwen2.5-7B2026.03 | 75.4 | |
| ICRLModel=Qwen2.5-3B2026.03 | 72.6 | |
| ZeroSearchModel=Qwen2.5-7B2026.03 | 65.2 | |
| ParallelSearchModel=Qwen2.5-7B2026.03 | 62.8 | |
| Search-R1Model=Qwen2.5-7B2026.03 | 61 | |
| Reject SamplingModel=Qwen2.5-7B2026.03 | 59.2 | |
| RAGModel=Qwen2.5-7B2026.03 | 58.5 | |
| ZeroSearchModel=Qwen2.5-3B2026.03 | 57.4 | |
| Search-R1Model=Qwen2.5-3B2026.03 | 54.5 | |
| RAGModel=Qwen2.5-3B2026.03 | 54.4 | |
| R1-baseModel=Qwen2.5-7B2026.03 | 53.9 | |
| R1-instructModel=Qwen2.5-7B2026.03 | 53.7 | |
| Reject SamplingModel=Qwen2.5-3B2026.03 | 48.8 | |
| IRCoTModel=Qwen2.5-7B2026.03 | 47.8 | |
| Search-o1Model=Qwen2.5-3B2026.03 | 47.2 | |
| R1-instructModel=Qwen2.5-3B2026.03 | 44.9 | |
| Search-o1Model=Qwen2.5-7B2026.03 | 44.3 | |
| DirectModel=Qwen2.5-7B2026.03 | 40.8 | |
| SFTModel=Qwen2.5-7B2026.03 | 35.4 | |
| IRCoTModel=Qwen2.5-3B2026.03 | 31.2 | |
| SFTModel=Qwen2.5-3B2026.03 | 29.2 | |
| DirectModel=Qwen2.5-3B2026.03 | 28.8 | |
| CoTModel=Qwen2.5-7B2026.03 | 18.5 | |
| CoTModel=Qwen2.5-3B2026.03 | 3.2 |