List your company

Mathematical Reasoning on GSM8k Base-7 (test)

49.9AccuracySFT+AdaBack
2.99615.17327.3539.527Jun 22, 2025

Evaluation results

MethodLinks
SFT+AdaBackModel Size=3B2025.0649.9
SFT+RLModel Size=3B2025.0645.4
AdaBackModel Size=3B2025.0643.9
SFT+AdaBackModel Size=1B2025.0624.5
AdaBackModel Size=1B2025.0618.4
SFT+RLModel Size=1B2025.0614.4
Base+RLModel Size=3B2025.064.9
Base+RLModel Size=1B2025.064.8