Mathematical Reasoning on GSM8K (test) (pass@1 accuracy)
78.3Pass@1 AccuracyLearnAlign
Evaluation Results
| Method | Links | |
|---|---|---|
| LearnAlignSelected Data Size=2,0002025.06 | 78.3 | |
| LearnAlignSelected Data Size=1,0002025.06 | 77.5 | |
| Qwen2.5-1.5B-Instruct-FULL2025.06 | 77 | |
| LIMRSelected Data Size=2,0002025.06 | 76.7 | |
| LearnAlignSelected Data Size=5002025.06 | 76.4 | |
| Token LengthSelected Data Size=1,0002025.06 | 76.2 | |
| LIMRSelected Data Size=5002025.06 | 76.2 | |
| LIMRSelected Data Size=1,0002025.06 | 76.1 | |
| IFDSelected Data Size=5002025.06 | 76 | |
| PPL-TopSelected Data Size=5002025.06 | 75.8 | |
| SelectITSelected Data Size=5002025.06 | 75.7 | |
| Random SamplingSelected Data Size=1,0002025.06 | 75.6 | |
| IFDSelected Data Size=1,0002025.06 | 75.6 | |
| Token LengthSelected Data Size=2,0002025.06 | 75.6 | |
| SelectITSelected Data Size=1,0002025.06 | 75.6 | |
| Random SamplingSelected Data Size=2,0002025.06 | 75.5 | |
| SelectITSelected Data Size=2,0002025.06 | 75.5 | |
| IFDSelected Data Size=2,0002025.06 | 75.4 | |
| PPL-TopSelected Data Size=2,0002025.06 | 75.2 | |
| Random SamplingSelected Data Size=5002025.06 | 75.1 | |
| PPL-MiddleSelected Data Size=1,0002025.06 | 75 | |
| LearnAlignSelected Data Size=1002025.06 | 74.8 | |
| PPL-MiddleSelected Data Size=5002025.06 | 74.7 | |
| PPL-TopSelected Data Size=1,0002025.06 | 74.6 | |
| Token LengthSelected Data Size=5002025.06 | 74.4 | |
| PPL-MiddleSelected Data Size=2,0002025.06 | 74.2 | |
| LIMRSelected Data Size=1002025.06 | 74.2 | |
| Random SamplingSelected Data Size=1002025.06 | 73.1 | |
| PPL-MiddleSelected Data Size=1002025.06 | 72.8 | |
| SelectITSelected Data Size=1002025.06 | 72.8 | |
| PPL-TopSelected Data Size=1002025.06 | 72.5 | |
| Token LengthSelected Data Size=1002025.06 | 72.3 | |
| IFDSelected Data Size=1002025.06 | 72 | |
| Qwen2.5-1.5B-Instruct2025.06 | 55.7 |