Mathematical Reasoning on AIME24, AIME25, MATH500, AMC23, Hmmt25, and Olympiad
61.67AIME 2024 ScoreIH-GRPO
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| IH-GRPOBackbone=Qwen3-8B, Tool=D2026.05 | 61.67 | 47.5 | 93.5 | 94.3 | 33.13 | 67.58 | 66.28 | |
| SimpleTIRBackbone=Qwen3-8B, Tool=D2026.05 | 58.33 | 44.69 | 92.52 | 91.02 | 31.35 | 64.57 | 63.75 | |
| Dr.GRPOBackbone=Qwen3-8B, Tool=D2026.05 | 57.5 | 41.35 | 92.7 | 90.55 | 31.98 | 66.5 | 63.43 | |
| Dr.GRPOBackbone=Qwen3-8B, Tool=C2026.05 | 56.35 | 41.15 | 92.5 | 89.14 | 31.04 | 65.07 | 62.54 | |
| DAPOBackbone=Qwen3-8B, Tool=D2026.05 | 56.15 | 43.23 | 92.35 | 91.56 | 32.81 | 64.5 | 63.43 | |
| EH-GRPOBackbone=Qwen3-8B, Tool=D2026.05 | 55.42 | 43.02 | 93.4 | 89.14 | 27.81 | 64.1 | 62.15 | |
| IH-GRPOBackbone=Qwen3-4B, Tool=D2026.05 | 55.31 | 45.83 | 91.45 | 90 | 35.1 | 64.88 | 63.76 | |
| SimpleTIRBackbone=Qwen3-8B, Tool=C2026.05 | 52.92 | 40.31 | 90.25 | 89.53 | 33.54 | 65.28 | 61.97 | |
| SimpleTIRBackbone=Qwen3-4B, Tool=D2026.05 | 51.98 | 41.25 | 91.33 | 89.14 | 32.29 | 63.6 | 61.6 | |
| Dr.GRPOBackbone=Qwen3-4B, Tool=D2026.05 | 51.67 | 43.65 | 88.7 | 86.02 | 35.73 | 63.17 | 61.49 | |
| DAPOBackbone=Qwen3-8B, Tool=C2026.05 | 51.56 | 38.75 | 88.52 | 88.67 | 31.46 | 64.07 | 60.51 | |
| DAPOBackbone=Qwen3-4B, Tool=D2026.05 | 50.1 | 40.83 | 91.1 | 89.3 | 30.63 | 63.15 | 60.85 | |
| Dr.GRPOBackbone=Qwen3-4B, Tool=C2026.05 | 49.69 | 42.71 | 90.3 | 87.11 | 30.42 | 63.9 | 60.69 | |
| DAPOBackbone=Qwen3-4B, Tool=C2026.05 | 49.38 | 42.6 | 91.15 | 88.83 | 31.87 | 64.57 | 61.4 | |
| SimpleTIRBackbone=Qwen3-4B, Tool=C2026.05 | 49.27 | 43.75 | 87.88 | 85.31 | 31.67 | 60.93 | 59.8 | |
| EH-GRPOBackbone=Qwen3-4B, Tool=D2026.05 | 41.77 | 35.31 | 90.25 | 83.28 | 26.98 | 61.92 | 56.59 | |
| STaRBackbone=Qwen3-8B, Tool=D2026.05 | 40.83 | 25 | 80.55 | 67.42 | 28.65 | 52.32 | 49.13 | |
| STaRBackbone=Qwen3-8B, Tool=C2026.05 | 38.85 | 27.29 | 79.85 | 67.73 | 26.98 | 53.55 | 49.04 | |
| BaseBackbone=Qwen3-8B, Tool=No2026.05 | 34.69 | 23.65 | 84.8 | 66.8 | 11.04 | 46.47 | 44.58 | |
| BaseBackbone=Qwen3-4B, Tool=No2026.05 | 34.17 | 24.69 | 86.35 | 70.7 | 12.5 | 47.33 | 45.96 | |
| STaRBackbone=Qwen3-4B, Tool=D2026.05 | 32.4 | 23.65 | 78.47 | 68.12 | 24.69 | 49.72 | 46.18 | |
| IH-GRPOBackbone=Qwen3-1.7B, Tool=D2026.05 | 31.15 | 27.5 | 83.8 | 66.87 | 18.96 | 51.55 | 46.64 | |
| Dr.GRPOBackbone=Qwen3-1.7B, Tool=D2026.05 | 29.37 | 26.25 | 82.2 | 62.66 | 16.98 | 51.13 | 44.77 | |
| SimpleTIRBackbone=Qwen3-1.7B, Tool=D2026.05 | 29.27 | 27.19 | 81.7 | 63.67 | 16.56 | 49.18 | 44.6 | |
| Dr.GRPOBackbone=Qwen3-1.7B, Tool=C2026.05 | 27.92 | 25.42 | 79.8 | 64.38 | 18.33 | 51.57 | 44.57 | |
| EH-GRPOBackbone=Qwen3-1.7B, Tool=D2026.05 | 27.81 | 22.6 | 82.02 | 64.06 | 17.4 | 50.17 | 44.01 | |
| SimpleTIRBackbone=Qwen3-1.7B, Tool=C2026.05 | 26.67 | 24.06 | 83.97 | 60.7 | 9.69 | 46.3 | 41.9 | |
| STaRBackbone=Qwen3-4B, Tool=C2026.05 | 26.35 | 22.6 | 73.77 | 61.09 | 20.62 | 48.93 | 42.23 | |
| DAPOBackbone=Qwen3-1.7B, Tool=C2026.05 | 25.73 | 26.56 | 83.75 | 67.73 | 14.06 | 50.77 | 44.77 | |
| DAPOBackbone=Qwen3-1.7B, Tool=D2026.05 | 24.9 | 23.54 | 78.78 | 60.08 | 15.83 | 46.55 | 41.61 | |
| BaseBackbone=Qwen3-1.7B, Tool=No2026.05 | 22.5 | 21.77 | 81.05 | 57.89 | 9.58 | 43.3 | 39.35 | |
| STaRBackbone=Qwen3-1.7B, Tool=D2026.05 | 11.67 | 11.46 | 58.82 | 38.12 | 8.44 | 32.55 | 26.84 | |
| STaRBackbone=Qwen3-1.7B, Tool=C2026.05 | 11.25 | 12.92 | 64.72 | 43.36 | 9.27 | 35.93 | 29.57 |