Mathematical Reasoning on Minerva (Acc@16)
26Acc@16Verify-Init
Evaluation Results
| Method | Links | |
|---|---|---|
| Verify-InitBackbone=Qwen2.5-7B-Instruct2026.02 | 26 | |
| Verify-AlterBackbone=Qwen2.5-7B-Instruct2026.02 | 26 | |
| GenerateBackbone=Qwen2.5-7B-Instruct2026.02 | 25.3 | |
| Mixed-TrainBackbone=Qwen2.5-7B-Instruct2026.02 | 24.6 | |
| Verify-AlterBackbone=Qwen2.5-3B-Instruct2026.02 | 18.7 | |
| Verify-InitBackbone=Qwen2.5-3B-Instruct2026.02 | 18.3 | |
| Mixed-TrainBackbone=Qwen2.5-3B-Instruct2026.02 | 17.6 | |
| GenerateBackbone=Qwen2.5-3B-Instruct2026.02 | 17 | |
| Verify-AlterBackbone=Qwen2.5-1.5B-Instruct2026.02 | 13.9 | |
| Verify-InitBackbone=Qwen2.5-1.5B-Instruct2026.02 | 13 | |
| Mixed-TrainBackbone=Qwen2.5-1.5B-Instruct2026.02 | 12.7 | |
| GenerateBackbone=Qwen2.5-1.5B-Instruct2026.02 | 12.4 |