Question Answering on HotpotQA In-distribution (test)
30.9AccuracyRLCR
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| RLCRBackbone=Mistral-7B-Instruct2026.05 | 30.9 | 0.2037 | 67.7 | |
| RLVRBackbone=Llama-3.1-8B-Instruct2026.05 | 30.3 | 0.1349 | 81.18 | |
| RDBackbone=Mistral-7B-Instruct2026.05 | 29.4 | 0.2615 | 72.79 | |
| CSRBackbone=Llama-3.1-8B-Instruct2026.05 | 27.46 | 0.0783 | 88.86 | |
| CSRBackbone=Mistral-7B-Instruct2026.05 | 27.32 | 0.0884 | 87.48 | |
| RLVRBackbone=Mistral-7B-Instruct2026.05 | 27.3 | 0.1314 | 81.12 | |
| RLCRBackbone=Llama-3.1-8B-Instruct2026.05 | 27.2 | 0.2137 | 66.16 | |
| RDBackbone=Llama-3.1-8B-Instruct2026.05 | 26.4 | 0.115 | 84.69 | |
| RLVRBackbone=Qwen2.5-7B-Instruct2026.05 | 26.36 | 0.1559 | 80.83 | |
| RDBackbone=Qwen2.5-7B-Instruct2026.05 | 25.5 | 0.1397 | 69.1 | |
| CSRBackbone=Qwen2.5-7B-Instruct2026.05 | 24.1 | 0.1159 | 89.5 | |
| BaseBackbone=Qwen2.5-7B-Instruct2026.05 | 23.99 | 0.1555 | 81.47 | |
| RLCRBackbone=Qwen2.5-7B-Instruct2026.05 | 21.9 | 0.2347 | 73.12 | |
| BaseBackbone=Mistral-7B-Instruct2026.05 | 18.37 | 0.1607 | 81.83 | |
| BaseBackbone=Llama-3.1-8B-Instruct2026.05 | 17.95 | 0.1821 | 76.66 |