Mathematical Reasoning on AIME 24 (Acc.@first, Acc.@final)
16.3Accuracy @ First AnswerCoVerRL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CoVerRLModel=Llama-3.2-3B-Instruct2026.03 | 16.3 | 16.9 | |
| CoVerRLModel=Qwen2.5-7B2026.03 | 14.6 | 17.1 | |
| TTRLModel=Llama-3.2-3B-Instruct2026.03 | 13.8 | 14 | |
| TTRLModel=Qwen2.5-7B2026.03 | 12.7 | 12.5 | |
| CoVerRLModel=Qwen3-1.7B-Base2026.03 | 9.8 | 10.6 | |
| TTRLModel=Qwen3-1.7B-Base2026.03 | 5.2 | 5.2 | |
| Base ModelModel=Qwen2.5-7B2026.03 | 5.2 | 6.5 | |
| Base ModelModel=Llama-3.2-3B-Instruct2026.03 | 4.6 | 5 | |
| Base ModelModel=Qwen3-1.7B-Base2026.03 | 3.8 | 3.3 |