List your company

Mathematical Reasoning on GSM8K (Strict-match Accuracy)

78.9Strict-match AccuracyS-SPPO
41.35651.10360.8570.597Jun 1, 2026

Evaluation results

MethodLinks
S-SPPOBackbone=Llama-3-8B, Iteration=Iter32026.0678.9
SPPOBackbone=Llama-3-8B, Iteration=Iter32026.0678.2
Llama-3-8B BaseBackbone=Llama-3-8B, Optimization Method=Base2026.0676.4
S-SPPOBackbone=Mistral-7B, Iteration=Iter32026.0644.2
SPPOBackbone=Mistral-7B, Iteration=Iter32026.0644
Mistral-7B BaseBackbone=Mistral-7B, Optimization Method=Base2026.0642.8