Mathematical Reasoning on AIME, AMC, MATH, Minerva, Olympiad (test)
21.4Avg@8 (AIME)AERO
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AEROModel Backbone=Qwen2.5-7B-Math2026.02 | 21.4 | 10.8 | 9,181 | 55.2 | 71.9 | 28.9 | 38.2 | 48 | 47.6 | 81.1 | 90.5 | 50.3 | 62.6 | 70 | |
| GRPOModel Backbone=Qwen2.5-7B-Math2026.02 | 20.8 | 21.5 | 17,334 | 54.3 | 71.4 | 26.7 | 37.2 | 46.9 | 47.6 | 72.1 | 90.2 | 49.1 | 60.7 | 68.6 | |
| DAPOModel Backbone=Qwen2.5-7B-Math2026.02 | 20.8 | 15.8 | 13,743 | 53.3 | 72 | 28.1 | 37.6 | 47.5 | 47.6 | 73.6 | 90.5 | 49.1 | 61.9 | 69.1 | |
| BaseModel Backbone=Qwen2.5-7B-Math2026.02 | 13.8 | — | — | 35.8 | 53.9 | 17.9 | 19.6 | 31.1 | 40 | 70 | 87.6 | 43.4 | 51 | 62.2 | |
| AEROModel Backbone=Qwen2.5-1.5B-Math2026.02 | 10.7 | 19.1 | 1,811 | 32.3 | 57.8 | 17.1 | 25.6 | 34.52 | 28.6 | 66 | 86.3 | 40.9 | 51.7 | 61.21 | |
| DAPOModel Backbone=Qwen2.5-1.5B-Math2026.02 | 10.1 | 26.1 | 2,722 | 33 | 56.1 | 17.1 | 25.7 | 34.05 | 33.3 | 64.2 | 84.3 | 40.9 | 52.6 | 61 | |
| GRPOModel Backbone=Qwen2.5-1.5B-Math2026.02 | 8.3 | 34.3 | 3,627 | 32.6 | 57.1 | 17.3 | 25.1 | 34.01 | 23.8 | 60.4 | 85.6 | 42.1 | 51.9 | 60.9 | |
| BaseModel Backbone=Qwen2.5-1.5B-Math2026.02 | 2.5 | — | — | 19.2 | 32.2 | 8.5 | 12.9 | 16.17 | 20 | 70 | 77.8 | 34.5 | 39.9 | 50.5 |