Mathematical Reasoning on AMC (Acc.@first, Acc.@final)
47.6Accuracy @ First StepCoVerRL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoVerRLModel=Qwen2.5-7B2026.03 | 47.6 | 49.2 | |
| TTRLModel=Qwen2.5-7B2026.03 | 42.2 | 42.2 | |
| CoVerRLModel=Qwen3-1.7B-Base2026.03 | 36 | 38.6 | |
| TTRLModel=Qwen3-1.7B-Base2026.03 | 31.1 | 30.9 | |
| CoVerRLModel=Llama-3.2-3B-Instruct2026.03 | 28.3 | 32.2 | |
| Base ModelModel=Qwen2.5-7B2026.03 | 25.5 | 26.4 | |
| Base ModelModel=Qwen3-1.7B-Base2026.03 | 24.6 | 24.5 | |
| TTRLModel=Llama-3.2-3B-Instruct2026.03 | 23.8 | 23.3 | |
| Base ModelModel=Llama-3.2-3B-Instruct2026.03 | 17 | 15.7 |