Mathematical Reasoning on AIME26 (Maj@32, Avg@32)
66.7Maj@32PAEC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PAECModel=Qwen3-4B-Base2026.06 | 66.7 | 43.2 | |
| GRPOModel=Qwen3-4B-Base2026.06 | 63.3 | 42.9 | |
| Global Entropy RegModel=Qwen3-4B-Base2026.06 | 53.3 | 40.3 | |
| BaseModel=Qwen3-4B-Base2026.06 | 50 | 38.4 | |
| PAECModel=Llama3.1-8B2026.06 | 6.7 | 1 | |
| GRPOModel=Llama3.1-8B2026.06 | 3.3 | 0.2 | |
| Global Entropy RegModel=Llama3.1-8B2026.06 | 3.3 | 0.5 | |
| BaseModel=Llama3.1-8B2026.06 | 0 | 0.1 |