MuggleMath: Assessing the Impact of Query and Response Augmentation on Math Reasoning

Benchmark Model Rank Results
math-word-problem-solving-on-mathMuggleMATH-70B#127Accuracy: 35.6Parameters (Billions): 70
math-word-problem-solving-on-mathMuggleMATH-13B#138Accuracy: 30.7Parameters (Billions): 13
math-word-problem-solving-on-mathMuggleMATH 7B#147Accuracy: 25.8Parameters (Billions): 7
arithmetic-reasoning-on-gsm8kMuggleMATH 70B#145Accuracy: 82.3Parameters (Billion): 70
arithmetic-reasoning-on-gsm8kMuggleMATH 13B#186Accuracy: 74Parameters (Billion): 13
arithmetic-reasoning-on-gsm8kMuggleMATH 7B#199Accuracy: 69.8Parameters (Billion): 7