Mathematical Reasoning on MATH500 (Acc.@first, Acc.@final)
76.8Accuracy @ First StepCoVerRL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoVerRLModel=Qwen2.5-7B2026.03 | 76.8 | 79.6 | |
| TTRLModel=Qwen2.5-7B2026.03 | 73.8 | 74.2 | |
| CoVerRLModel=Qwen3-1.7B-Base2026.03 | 69 | 71.9 | |
| TTRLModel=Qwen3-1.7B-Base2026.03 | 65.1 | 65 | |
| CoVerRLModel=Llama-3.2-3B-Instruct2026.03 | 55.9 | 59.3 | |
| Base ModelModel=Qwen3-1.7B-Base2026.03 | 53.5 | 53.3 | |
| TTRLModel=Llama-3.2-3B-Instruct2026.03 | 52.6 | 52.2 | |
| Base ModelModel=Qwen2.5-7B2026.03 | 50.1 | 51.4 | |
| Base ModelModel=Llama-3.2-3B-Instruct2026.03 | 42.7 | 41 |