Mathematical Reasoning on AMC 23 (pass@1, pass@32)
75.6Pass@1AER
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| AERModel Backbone=Qwen3-8B-Base2025.10 | 75.6 | 94.8 | |
| Clip-CovModel Backbone=Qwen3-8B-Base2025.10 | 74.5 | 94.1 | |
| AERModel Backbone=Qwen3-4B-Base2025.10 | 70.2 | 94.8 | |
| Clip-HigherModel Backbone=Qwen3-8B-Base2025.10 | 68.5 | 92.7 | |
| Clip-CovModel Backbone=Qwen3-4B-Base2025.10 | 65.9 | 92.4 | |
| KL-CovModel Backbone=Qwen3-8B-Base2025.10 | 65.3 | 92.1 | |
| GRPOModel Backbone=Qwen3-8B-Base2025.10 | 62.8 | 88.4 | |
| Ent-AdvModel Backbone=Qwen3-8B-Base2025.10 | 62.5 | 80.6 | |
| GRPOModel Backbone=Qwen3-4B-Base2025.10 | 59.9 | 87.8 | |
| Clip-HigherModel Backbone=Qwen3-4B-Base2025.10 | 59.6 | 89.7 | |
| KL-CovModel Backbone=Qwen3-4B-Base2025.10 | 57.2 | 84.9 | |
| Ent-AdvModel Backbone=Qwen3-4B-Base2025.10 | 49.2 | 86.7 | |
| BaseModel Backbone=Qwen3-8B-Base2025.10 | 45 | 88.7 | |
| BaseModel Backbone=Qwen3-4B-Base2025.10 | 34.6 | 86.7 |