Mathematical Reasoning on AMC 2023 (avg@8)
70.3Avg@8 ScoreSHEAR
Evaluation Results
| Method | Links | |
|---|---|---|
| SHEARBackbone=Qwen2.5-Math-7B2026.04 | 70.3 | |
| PRM(PURE)Backbone=Qwen2.5-Math-7B2026.04 | 70 | |
| PRM(Reshape adv.)Backbone=Qwen2.5-Math-7B2026.04 | 66.8 | |
| GRPOBackbone=Qwen2.5-Math-7B2026.04 | 66.5 | |
| Entropy adv.Backbone=Qwen2.5-Math-7B2026.04 | 66 | |
| Entropy adv.Backbone=Qwen2.5-14B-Base2026.04 | 64.8 | |
| GRPOBackbone=Qwen2.5-14B-Base2026.04 | 62.4 | |
| SHEARBackbone=Qwen2.5-14B-Base2026.04 | 62.2 | |
| PRM(Reshape adv.)Backbone=Qwen2.5-14B-Base2026.04 | 58.5 | |
| PRM(PURE)Backbone=Qwen2.5-14B-Base2026.04 | 49.2 | |
| SHEARBackbone=Llama3.1-8B-Instruct2026.04 | 37.2 | |
| PRM(Reshape adv.)Backbone=Llama3.1-8B-Instruct2026.04 | 33.4 | |
| Entropy adv.Backbone=Llama3.1-8B-Instruct2026.04 | 33.1 | |
| GRPOBackbone=Llama3.1-8B-Instruct2026.04 | 32.7 | |
| PRM(PURE)Backbone=Llama3.1-8B-Instruct2026.04 | 27.9 |