Scaling Relationship on Learning Mathematical Reasoning with Large Language Models

Benchmark Model Rank Results
arithmetic-reasoning-on-gsm8kRFT 70B#209Accuracy: 64.8Parameters (Billion): 79
arithmetic-reasoning-on-gsm8kRFT 13B#225Accuracy: 55.3Parameters (Billion): 13
arithmetic-reasoning-on-gsm8kRFT 7B#236Accuracy: 51.2Parameters (Billion): 7