Mathematical Reasoning on AIME24, Math500, Olympiad, AMC, and Minerva
60.58AIME24 ScoreopenPangu-7B + DARS
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| openPangu-7B + DARSHardware=Huawei Ascend NPU, Backbone=openPangu-7B2025.08 | 60.58 | 93.95 | 67.96 | 81.06 | 43.4 | 72.56 | 86.59 | |
| openPangu-7BHardware=Huawei Ascend NPU, Backbone=openPangu-7B2025.08 | 36.99 | 82.71 | 48.82 | 51.61 | 37 | 57.54 | 79.92 | |
| CASPO+CaTBase Model=Qwen2.5-Math-7B2026.05 | 26.7 | 81.9 | 46.9 | 67.5 | 40.5 | 52.7 | — | |
| CASPO+CaTBase Model=Qwen2.5-7B-Instruct2026.05 | 26.7 | 87.7 | 51.7 | 67.5 | 47.1 | 56.1 | — | |
| DPO-VPBase Model=Qwen2.5-Math-7B2026.05 | 23.3 | 74.8 | 36.9 | 60 | 35.3 | 46.1 | — | |
| CASPOBase Model=Qwen2.5-Math-7B2026.05 | 23.3 | 76.6 | 43.8 | 62.5 | 37.8 | 48.8 | — | |
| Simple-RL-ZeroBase Model=Qwen2.5-Math-7B2026.05 | 20 | 78 | 36.6 | 57.5 | 33.1 | 45 | — | |
| PURE-VRBase Model=Qwen2.5-Math-7B2026.05 | 20 | 79.8 | 41.9 | 57.5 | 36.8 | 47.5 | — | |
| DPO-VPBase Model=Qwen2.5-7B-Instruct2026.05 | 20 | 79.8 | 46.2 | 60 | 42.5 | 49.1 | — | |
| CASPOBase Model=Qwen2.5-7B-Instruct2026.05 | 20 | 82 | 48.3 | 62.5 | 44 | 50.6 | — | |
| CASPO+CaTBase Model=Llama-3.1-8B-Instruct2026.05 | 20 | 59.1 | 29.5 | 40 | 16.7 | 33.1 | — | |
| Qwen2.5-Math-7BBase Model=Qwen2.5-Math-7B2026.05 | 16.7 | 64.8 | 25.6 | 37.5 | 15.4 | 32 | — | |
| GRPOBase Model=Qwen2.5-Math-7B2026.05 | 16.7 | 76.2 | 38.1 | 55 | 32.7 | 43.7 | — | |
| Simple-RL-ZeroBase Model=Qwen2.5-7B-Instruct2026.05 | 16.7 | 80.2 | 45.8 | 57.5 | 41.5 | 47.8 | — | |
| PURE-VRBase Model=Qwen2.5-7B-Instruct2026.05 | 16.7 | 81.5 | 47.5 | 57.5 | 43 | 48.9 | — | |
| Qwen2.5-7B-InstructBase Model=Qwen2.5-7B-Instruct2026.05 | 13.3 | 76.2 | 43 | 52.5 | 37.6 | 44.4 | — | |
| GRPOBase Model=Qwen2.5-7B-Instruct2026.05 | 13.3 | 79 | 46.5 | 55 | 41 | 46.6 | — | |
| DPO-VPBase Model=Llama-3.1-8B-Instruct2026.05 | 13.3 | 54.8 | 27.2 | 32.5 | 16.5 | 28.8 | — | |
| CASPOBase Model=Llama-3.1-8B-Instruct2026.05 | 13.3 | 55.2 | 27.6 | 35 | 15.6 | 29.1 | — | |
| Simple-RL-ZeroBase Model=Llama-3.1-8B-Instruct2026.05 | 10 | 53.2 | 25.6 | 30 | 15.5 | 26.9 | — | |
| PURE-VRBase Model=Llama-3.1-8B-Instruct2026.05 | 10 | 54 | 26.8 | 32.5 | 16 | 27.6 | — | |
| Llama-3.1-8B-InstructBase Model=Llama-3.1-8B-Instruct2026.05 | 6.7 | 49.6 | 23.5 | 27.5 | 13.2 | 24.1 | — | |
| GRPOBase Model=Llama-3.1-8B-Instruct2026.05 | 6.7 | 52 | 25 | 30 | 15 | 25.5 | — |