Mathematical Reasoning on MATH500 (Maj@8, Avg@8)
90Maj@8PAEC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PAECModel=Qwen3-4B-Base2026.06 | 90 | 86.8 | |
| Global Entropy RegModel=Qwen3-4B-Base2026.06 | 88.6 | 86.4 | |
| GRPOModel=Qwen3-4B-Base2026.06 | 88.2 | 86.4 | |
| BaseModel=Qwen3-4B-Base2026.06 | 86.8 | 84.2 | |
| PAECModel=Llama3.1-8B2026.06 | 26.4 | 21.8 | |
| GRPOModel=Llama3.1-8B2026.06 | 25.6 | 21.3 | |
| Global Entropy RegModel=Llama3.1-8B2026.06 | 20.6 | 18.4 | |
| BaseModel=Llama3.1-8B2026.06 | 9.8 | 4.6 |