Mathematical Reasoning on AMC 23 (Acc@16, Tokens)
65.3Accuracy@16Generate
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GenerateModel=Qwen2.5-7B-Instruct2026.02 | 65.3 | 3,967 | |
| Self-VerifyModel=Qwen2.5-7B-Instruct2026.02 | 59.7 | 1,194 | |
| GenerateModel=Qwen2.5-3B-Instruct2026.02 | 50.9 | 2,754 | |
| Self-VerifyModel=Qwen2.5-3B-Instruct2026.02 | 46.7 | 1,825 | |
| Self-VerifyModel=Qwen2.5-1.5B-Instruct2026.02 | 33 | 1,309 | |
| GenerateModel=Qwen2.5-1.5B-Instruct2026.02 | 30.5 | 1,402 |