Improving Language Model Reasoning with Self-motivated Learning

Yunlong Feng, Yang Xu, Libo Qin +2

Large-scale high-quality training data is important for improving the performance of models. After trained with data that has rationales (reasoning steps), models gain reasoning ca…