Mathematical Reasoning on AIME 25 (Pass@1, Token count)
72.2Pass@1 AccuracyRED
Evaluation results
| Method | Links | ||
|---|---|---|---|
| REDModel=Qwen3-32B-thinking2026.04 | 72.2 | 5,908 | |
| AlphaOneModel=Qwen3-32B-thinking2026.04 | 71.1 | 8,569 | |
| S-GRPOModel=Qwen3-32B-thinking2026.04 | 71.1 | 6,389 | |
| GRPOModel=Qwen3-32B-thinking2026.04 | 70 | 12,487 | |
| VanillaModel=Qwen3-32B-thinking2026.04 | 68.9 | 11,589 | |
| DASTModel=Qwen3-32B-thinking2026.04 | 68.9 | 6,504 | |
| DEERModel=Qwen3-32B-thinking2026.04 | 66.7 | 8,417 | |
| RL + LPModel=Qwen3-32B-thinking2026.04 | 66.7 | 6,854 | |
| Think or NotModel=Qwen3-32B-thinking2026.04 | 65.6 | 6,772 | |
| AlphaOneModel=Qwen3-8B-thinking2026.04 | 63.3 | 8,607 | |
| REDModel=Qwen3-8B-thinking2026.04 | 63.3 | 6,203 | |
| REDModel=DpSk-R1-Distill-Qwen-32B2026.04 | 63.3 | 5,018 | |
| Think or NotModel=Qwen3-8B-thinking2026.04 | 62.2 | 6,674 | |
| GRPOModel=Qwen3-8B-thinking2026.04 | 62.2 | 11,098 | |
| RL + LPModel=DpSk-R1-Distill-Qwen-32B2026.04 | 62.2 | 6,124 | |
| GRPOModel=DpSk-R1-Distill-Qwen-32B2026.04 | 62.2 | 8,990 | |
| VanillaModel=Qwen3-8B-thinking2026.04 | 61.1 | 12,490 | |
| S-GRPOModel=Qwen3-8B-thinking2026.04 | 61.1 | 7,331 | |
| AlphaOneModel=DpSk-R1-Distill-Qwen-32B2026.04 | 61.1 | 8,994 | |
| DASTModel=Qwen3-8B-thinking2026.04 | 60 | 6,055 | |
| S-GRPOModel=DpSk-R1-Distill-Qwen-32B2026.04 | 60 | 5,334 | |
| DEERModel=Qwen3-8B-thinking2026.04 | 58.9 | 9,271 | |
| RL + LPModel=Qwen3-8B-thinking2026.04 | 58.9 | 7,639 | |
| VanillaModel=DpSk-R1-Distill-Qwen-32B2026.04 | 58.9 | 8,906 | |
| Think or NotModel=DpSk-R1-Distill-Qwen-32B2026.04 | 58.9 | 4,711 | |
| DEERModel=DpSk-R1-Distill-Qwen-32B2026.04 | 57.8 | 8,008 | |
| DASTModel=DpSk-R1-Distill-Qwen-32B2026.04 | 54.4 | 6,647 | |
| GRPOModel=DeepSeek-R1-Distill-Llama-70B2026.04 | 51.1 | 8,715 | |
| AlphaOneModel=DeepSeek-R1-Distill-Llama-70B2026.04 | 50 | 8,689 | |
| REDModel=DeepSeek-R1-Distill-Llama-70B2026.04 | 50 | 4,303 | |
| DEERModel=DeepSeek-R1-Distill-Llama-70B2026.04 | 48.9 | 7,478 | |
| S-GRPOModel=DeepSeek-R1-Distill-Llama-70B2026.04 | 48.9 | 5,896 | |
| REDModel=DpSk-R1-Distill-Qwen-7B2026.04 | 47.8 | 5,690 | |
| VanillaModel=DeepSeek-R1-Distill-Llama-70B2026.04 | 47.8 | 8,909 | |
| Think or NotModel=DeepSeek-R1-Distill-Llama-70B2026.04 | 46.7 | 4,414 | |
| GRPOModel=DpSk-R1-Distill-Qwen-7B2026.04 | 45.6 | 12,006 | |
| DASTModel=DeepSeek-R1-Distill-Llama-70B2026.04 | 45.6 | 5,932 | |
| RL + LPModel=DeepSeek-R1-Distill-Llama-70B2026.04 | 45.6 | 5,967 | |
| AlphaOneModel=DpSk-R1-Distill-Qwen-7B2026.04 | 44.4 | 8,921 | |
| S-GRPOModel=DpSk-R1-Distill-Qwen-7B2026.04 | 44.4 | 5,794 | |
| VanillaModel=DpSk-R1-Distill-Qwen-7B2026.04 | 43.3 | 11,454 | |
| RL + LPModel=DpSk-R1-Distill-Qwen-7B2026.04 | 43.3 | 6,255 | |
| DEERModel=DpSk-R1-Distill-Qwen-7B2026.04 | 42.2 | 8,261 | |
| DASTModel=DpSk-R1-Distill-Qwen-7B2026.04 | 41.1 | 7,904 | |
| Think or NotModel=DpSk-R1-Distill-Qwen-7B2026.04 | 38.9 | 4,760 | |
| AlphaOneModel=DeepSeek-R1-Distill-Llama-8B2026.04 | 34.4 | 9,005 | |
| REDModel=DeepSeek-R1-Distill-Llama-8B2026.04 | 34.4 | 5,039 | |
| DASTModel=DeepSeek-R1-Distill-Llama-8B2026.04 | 32.2 | 8,438 | |
| DEERModel=DeepSeek-R1-Distill-Llama-8B2026.04 | 31.1 | 8,936 | |
| S-GRPOModel=DeepSeek-R1-Distill-Llama-8B2026.04 | 31.1 | 5,426 | |
| Think or NotModel=DeepSeek-R1-Distill-Llama-8B2026.04 | 30 | 7,158 | |
| RL + LPModel=DeepSeek-R1-Distill-Llama-8B2026.04 | 30 | 5,897 | |
| GRPOModel=DeepSeek-R1-Distill-Llama-8B2026.04 | 30 | 11,987 | |
| VanillaModel=DeepSeek-R1-Distill-Llama-8B2026.04 | 28.9 | 11,548.2 |