Mathematical Reasoning on MATH-500 (maj@16, pass@16)
97.2Maj@16PPC
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| PPCBackbone=Qwen3-4B2026.05 | 97.2 | 99.4 | |
| GRPOBackbone=Qwen3-4B2026.05 | 96.6 | 98.8 | |
| Prompt-OnlyBackbone=Qwen3-4B2026.05 | 96.4 | 98.2 | |
| Plan-TuningBackbone=Qwen3-4B2026.05 | 96.2 | 98.8 | |
| BaseBackbone=Qwen3-4B2026.05 | 96 | 98.2 | |
| PTA-GRPOBackbone=Qwen3-4B2026.05 | 95.8 | 98.6 | |
| PPCBackbone=Qwen2.5-Math-7B2026.05 | 88.6 | 95 | |
| PTA-GRPOBackbone=Qwen2.5-Math-7B2026.05 | 87.8 | 92.8 | |
| Prompt-OnlyBackbone=Qwen2.5-Math-7B2026.05 | 87.6 | 91.8 | |
| Plan-TuningBackbone=Qwen2.5-Math-7B2026.05 | 87.6 | 93.6 | |
| BaseBackbone=Qwen2.5-Math-7B2026.05 | 87 | 90.8 | |
| GRPOBackbone=Qwen2.5-Math-7B2026.05 | 86.8 | 93.4 | |
| PPCBackbone=Qwen2.5-7B2026.05 | 84.8 | 92.2 | |
| Plan-TuningBackbone=Qwen2.5-7B2026.05 | 84 | 88.8 | |
| GRPOBackbone=Qwen2.5-7B2026.05 | 83.8 | 90.8 | |
| PTA-GRPOBackbone=Qwen2.5-7B2026.05 | 81.4 | 89.4 | |
| BaseBackbone=Qwen2.5-7B2026.05 | 80.6 | 89 | |
| Prompt-OnlyBackbone=Qwen2.5-7B2026.05 | 80.4 | 90 | |
| PPCBackbone=Llama3.1-8B2026.05 | 65.8 | 85 | |
| PTA-GRPOBackbone=Llama3.1-8B2026.05 | 60.6 | 82.8 | |
| Plan-TuningBackbone=Llama3.1-8B2026.05 | 60.2 | 83 | |
| GRPOBackbone=Llama3.1-8B2026.05 | 59.6 | 83.4 | |
| Prompt-OnlyBackbone=Llama3.1-8B2026.05 | 58 | 80 | |
| BaseBackbone=Llama3.1-8B2026.05 | 55.2 | 82 |