Mathematical Reasoning on OlympiadBench (avg@1)
43Avg Score (avg@1)STEER
Evaluation Results
| Method | Links | |
|---|---|---|
| STEERBackbone=Qwen2.5-Math-7B2025.10 | 43 | |
| Entro. Adv.Backbone=Qwen2.5-Math-7B2025.10 | 42.8 | |
| Clip–CovBackbone=Qwen2.5-Math-7B2025.10 | 41.3 | |
| OPOBackbone=Qwen2.5-Math-7B2025.10 | 41 | |
| GRPOBackbone=Qwen2.5-Math-7B2025.10 | 40.9 | |
| GRPO w/ Fork TokensBackbone=Qwen2.5-Math-7B2025.10 | 40.9 | |
| W-REINFORCEBackbone=Qwen2.5-Math-7B2025.10 | 40.9 | |
| SimpleRL-ZooBackbone=Qwen2.5-Math-7B2025.10 | 40.8 | |
| Eurus-PRIMEBackbone=Qwen2.5-Math-7B2025.10 | 40.6 | |
| KL–CovBackbone=Qwen2.5-Math-7B2025.10 | 39.4 | |
| GRPO w/ clip-highBackbone=Qwen2.5-Math-7B2025.10 | 39.3 | |
| GRPO w/ Entro. LossBackbone=Qwen2.5-Math-7B2025.10 | 37.9 | |
| Qwen2.5-Math-7BBackbone=Qwen2.5-Math-7B2025.10 | 13.8 |