Mathematical Reasoning on MATH 500 (pass@8, avg@8)
89.38Average Score (k=8)DAPO w/ PR-FKL
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DAPO w/ PR-FKLModel Backbone=Qwen3-8B-Base2026.05 | 89.38 | 96.2 | |
| DAPO w/ RRModel Backbone=Qwen3-8B-Base2026.05 | 89.33 | 96.8 | |
| DAPO w/ PR-RKLModel Backbone=Qwen3-8B-Base2026.05 | 88.58 | 95.6 | |
| DAPOModel Backbone=Qwen3-8B-Base2026.05 | 87.65 | 96 | |
| DAPO w/ PR-FKLModel Backbone=Qwen3-4B-Base2026.05 | 86.48 | 94.4 | |
| DAPO w/ PR-RKLModel Backbone=Qwen3-4B-Base2026.05 | 86.2 | 94.2 | |
| DAPOModel Backbone=Qwen3-4B-Base2026.05 | 86.1 | 94.2 | |
| DAPO w/ RRModel Backbone=Qwen3-4B-Base2026.05 | 85.98 | 95 | |
| BaseModel Backbone=Qwen3-4B-Base2026.05 | 62.38 | 89 | |
| BaseModel Backbone=Qwen3-8B-Base2026.05 | 60.93 | 89.8 |