Mathematical Reasoning on BeyondAIME (pass@1, pass@64)
8.3Pass@1CurveRL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CurveRLBackbone=Qwen3-4B-Base2026.05 | 8.3 | 40.4 | |
| MaxRLBackbone=Qwen3-4B-Base2026.05 | 8.2 | 32.5 | |
| GRPOBackbone=Qwen3-4B-Base2026.05 | 8 | 29 | |
| BaseBackbone=Qwen3-4B-Base2026.05 | 4 | 23.8 | |
| CurveRLBackbone=Qwen3-1.7B-Base2026.05 | 1.4 | 23.6 | |
| MaxRLBackbone=Qwen3-1.7B-Base2026.05 | 1.3 | 21.8 | |
| BaseBackbone=Qwen3-1.7B-Base2026.05 | 1.1 | 16.9 | |
| GRPOBackbone=Qwen3-1.7B-Base2026.05 | 1.1 | 15.6 |