MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Models

Benchmark Model Rank Results
arithmetic-reasoning-on-gsm8kMetaMath 70B#144Accuracy: 82.3Parameters (Billion): 70
arithmetic-reasoning-on-gsm8kMetaMath-Mistral-7B#172Accuracy: 77.7Parameters (Billion): 7
arithmetic-reasoning-on-gsm8kMetaMath 13B#198Accuracy: 71.0Parameters (Billion): 13
arithmetic-reasoning-on-gsm8kMetaMath 7B#206Accuracy: 66.4Parameters (Billion): 7
math-word-problem-solving-on-mathMetaMath 70B#146Accuracy: 26.0Parameters (Billions): 70
math-word-problem-solving-on-mathMetaMath 13B#156Accuracy: 22.5Parameters (Billions): 13
math-word-problem-solving-on-mathMetaMath 7B#162Accuracy: 19.4Parameters (Billions): 7