Mathematical Reasoning on GSM8K (Strict-match Accuracy)
78.9Strict-match AccuracyS-SPPO
Evaluation Results
| Method | Links | |
|---|---|---|
| S-SPPOBackbone=Llama-3-8B, Iteration=Iter32026.06 | 78.9 | |
| SPPOBackbone=Llama-3-8B, Iteration=Iter32026.06 | 78.2 | |
| Llama-3-8B BaseBackbone=Llama-3-8B, Optimization Method=Base2026.06 | 76.4 | |
| S-SPPOBackbone=Mistral-7B, Iteration=Iter32026.06 | 44.2 | |
| SPPOBackbone=Mistral-7B, Iteration=Iter32026.06 | 44 | |
| Mistral-7B BaseBackbone=Mistral-7B, Optimization Method=Base2026.06 | 42.8 |