Mathematical Reasoning on Combined Benchmarks AVG
52.03Average AccuracyOGER
Evaluation Results
| Method | Links | |
|---|---|---|
| OGERBackbone=Qwen2.5-Math-7B2026.04 | 52.03 | |
| LuffyBackbone=Qwen2.5-Math-7B2026.04 | 48.66 | |
| GRPOBackbone=Qwen2.5-Math-7B2026.04 | 39.12 | |
| OGERBackbone=Qwen2.5-Math-1.5B2026.04 | 36.77 | |
| LuffyBackbone=Qwen2.5-Math-1.5B2026.04 | 35.25 | |
| GRPOBackbone=Qwen2.5-Math-1.5B2026.04 | 28.69 | |
| BaseBackbone=Qwen2.5-Math-7B2026.04 | 28.61 | |
| BaseBackbone=Qwen2.5-Math-1.5B2026.04 | 13.95 |