Math Reasoning on AIME 2024 (Pass@k Suite)
63.3Pass@1Merged (α=0.2)
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Merged (α=0.2)Model variant=Merged (α=0.2), Interpolation Weight=0.22026.04 | 63.3 | — | — | — | — | — | — | — | — | |
| Qwen3-4B-Instruct-2507Model variant=Base Model2026.04 | 60 | — | — | — | — | — | — | — | — | |
| SFT No-TestModel variant=SFT No-Test2026.04 | 40 | — | — | — | — | — | — | — | — | |
| THRBase Model=Qwen2.5-Math-7B, p=< 02025.10 | 17.3 | 22.6 | 28.5 | 35.5 | 43.2 | 51.3 | 58.8 | 66.3 | 73.3 | |
| THRBase Model=Qwen2.5-Math-7B2025.10 | 15.7 | 21.3 | 27.3 | 34.7 | 43.2 | 51.4 | 58.4 | 63.6 | 66.7 | |
| GRPOBase Model=Qwen2.5-Math-7B2025.10 | 14.4 | 20.7 | 27.5 | 34.7 | 42 | 49.6 | 58.1 | 67.3 | 76.7 | |
| THRBase Model=Qwen2.5-Math-7B, p=> 02025.10 | 13.6 | 19 | 24.9 | 31.8 | 39.9 | 48.8 | 57.3 | 64.2 | 70 | |
| THRBase Model=Qwen2.5-Math-1.5B, p=< 02025.10 | 11.9 | 18.2 | 24.9 | 31.2 | 37.9 | 45.3 | 52.9 | 61.2 | 70 | |
| GRPOBase Model=Qwen2.5-Math-1.5B2025.10 | 11.4 | 17.7 | 24.3 | 30.5 | 36.7 | 43.4 | 50 | 56 | 63.3 | |
| THRBase Model=Qwen2.5-Math-1.5B2025.10 | 10.6 | 16.7 | 23.4 | 30.2 | 37.2 | 44.8 | 51.9 | 58.5 | 63.3 | |
| THRBase Model=Qwen2.5-Math-1.5B, p=> 02025.10 | 8.4 | 13.6 | 20 | 27 | 34.7 | 43.1 | 50.8 | 57.6 | 63.3 | |
| BaseBase Model=Qwen2.5-Math-7B2025.10 | 7.5 | 13.5 | 22 | 32 | 41 | 47.9 | 53.7 | 59.4 | 66.7 | |
| BaseBase Model=Qwen2.5-Math-1.5B2025.10 | 3.3 | 6.3 | 11.3 | 18.5 | 27.4 | 36.4 | 44.3 | 49.6 | 53.3 | |
| GRPOBase Model=Qwen2.5-0.5B-Instruct2025.10 | 0.4 | 0.8 | 1.5 | 2.9 | 5.4 | 10 | 17.2 | 27.3 | 36.7 | |
| THRBase Model=Qwen2.5-0.5B-Instruct2025.10 | 0.4 | 0.7 | 1.5 | 2.9 | 5.4 | 9.7 | 15.7 | 22 | 26.7 | |
| THRBase Model=Qwen2.5-0.5B-Instruct, p=< 02025.10 | 0.4 | 0.8 | 1.5 | 2.9 | 5.4 | 9.4 | 14.9 | 21.5 | 30 | |
| THRBase Model=Qwen2.5-0.5B-Instruct, p=> 02025.10 | 0.4 | 0.7 | 1.4 | 2.6 | 4.7 | 8.1 | 12.9 | 19.9 | 30 | |
| BaseBase Model=Qwen2.5-0.5B-Instruct2025.10 | 0.1 | 0.2 | 0.4 | 0.8 | 1.6 | 3.1 | 5.6 | 9.8 | 16.7 |