adalaw/MetaMath-Mistral-7B-MFT

Introduction

The model is trained with Masked Thought Fine-Tuning (MFT), a simple variant of standard Supervised Fine-Tuning (SFT). You can refer to our code and paper below.

Links

• Code: https://github.com/ChangyuChen347/MaskedThought
• Paper: https://arxiv.org/abs/2403.02178

Results

We test it with the scripts provided in MetaMath.

Model	GSM8K	MATH
adalaw/MetaMath-Mistral-7B-MFT	79.90	29.0
meta-math/MetaMath-Mistral-7B-SFT	77.70	28.2