Mathematical Reasoning on AMC23 (Avg@32 Accuracy)
90.5Avg@32 AccuracyDynamic Sampling (Oracle)
Evaluation results
| Method | Links | |
|---|---|---|
| Dynamic Sampling (Oracle)Backbone=DeepSeek-R1-Distill-7B, Strategy=DS, Runtime=77h2026.02 | 90.5 | |
| GPSBackbone=DeepSeek-R1-Distill-7B, Strategy=GPS, Runtime=49h2026.02 | 90.5 | |
| PCLBackbone=DeepSeek-R1-Distill-7B, Strategy=PCL, Runtime=50h2026.02 | 90.1 | |
| GRESOBackbone=DeepSeek-R1-Distill-7B, Strategy=GRESO, Runtime=53h2026.02 | 89.7 | |
| Uniform SamplingBackbone=DeepSeek-R1-Distill-7B, Strategy=Uniform, Runtime=40h2026.02 | 89.5 | |
| MoPPSBackbone=DeepSeek-R1-Distill-7B, Strategy=MoPPS, Runtime=42h2026.02 | 89.5 | |
| Dynamic Sampling (Oracle)Backbone=DeepSeek-R1-Distill-1.5B, Strategy=DS, Runtime=30h2026.02 | 79.2 | |
| GPSBackbone=DeepSeek-R1-Distill-1.5B, Strategy=GPS, Runtime=16h2026.02 | 78.1 | |
| MoPPSBackbone=DeepSeek-R1-Distill-1.5B, Strategy=MoPPS, Runtime=17h2026.02 | 77.8 | |
| PCLBackbone=DeepSeek-R1-Distill-1.5B, Strategy=PCL, Runtime=17h2026.02 | 77.7 | |
| GRESOBackbone=DeepSeek-R1-Distill-1.5B, Strategy=GRESO, Runtime=27h2026.02 | 77.1 | |
| DeepSeek-R1-Distill-7BBackbone=DeepSeek-R1-Distill-7B, Strategy=Base2026.02 | 76.5 | |
| Uniform SamplingBackbone=DeepSeek-R1-Distill-1.5B, Strategy=Uniform, Runtime=16h2026.02 | 76.1 | |
| DeepSeek-R1-Distill-1.5BBackbone=DeepSeek-R1-Distill-1.5B, Strategy=Base2026.02 | 57.7 |