Math Reasoning on HMMT 11/25 (pass@1, pass@64)
8.1Pass@1MaxRL
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MaxRLModel Backbone=Qwen3-4B-Base2026.05 | 8.1 | 35.6 | |
| CurveRLModel Backbone=Qwen3-4B-Base2026.05 | 7.2 | 39.1 | |
| GRPOModel Backbone=Qwen3-4B-Base2026.05 | 5.7 | 27.8 | |
| CurveRLModel Backbone=Qwen3-1.7B-Base2026.05 | 3.9 | 20.6 | |
| MaxRLModel Backbone=Qwen3-1.7B-Base2026.05 | 3.7 | 18.4 | |
| BaseModel Backbone=Qwen3-4B-Base2026.05 | 3.5 | 21.3 | |
| GRPOModel Backbone=Qwen3-1.7B-Base2026.05 | 3 | 14.7 | |
| BaseModel Backbone=Qwen3-1.7B-Base2026.05 | 2.4 | 18.7 |