Mathematical Reasoning on Olympiad (Accuracy, Inference Latency)
64.76AccuracyHDPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HDPOBackbone=Qwen2.5-Math-7B2026.06 | 64.76 | 0.22 | |
| Critique-GRPOBackbone=Qwen2.5-Math-7B2026.06 | 61.46 | 0.26 | |
| LUFFYBackbone=Qwen2.5-Math-7B2026.06 | 56.09 | 0.18 | |
| GRPOBackbone=Qwen2.5-Math-7B2026.06 | 53.68 | 0.18 | |
| OatBackbone=Qwen2.5-Math-7B2026.06 | 50.82 | 0.1 | |
| SimpleRLBackbone=Qwen2.5-Math-7B2026.06 | 49.84 | 0.12 | |
| EDGE-GRPOBackbone=Qwen2.5-Math-7B2026.06 | 48.79 | 0.08 |