MuMath: Multi-perspective Data Augmentation for Mathematical Reasoning in Large Language Models

Anonymous

TL;DR: Broadening the scope of augmentation methods to construct a multi-perspective augmentation dataset for mathematics, termed MuMath Dataset

Abstract:

Recently, the tool-use Large Language Models (LLMs) that integrate with external Python interpreters have significantly enhanced mathematical reasoning capabilities for open-source LLMs. However, these models fall short in demonstrating the calculation process, which compromises user-friendliness and understanding of problem-solving steps. Conversely, while tool-free methods offer a clear display of the problem-solving process, their accuracy leaves room for improvement.
These tool-free methods typically employ a somewhat narrow range of augmentation techniques such as rephrasing and complexity enhancement to boost performance. In response to this issue, we have amalgamated and further refined these strengths while broadening the scope of augmentation methods to construct a multi-perspective augmentation dataset for mathematics---termed MuMath Dataset.
Subsequently, we finetune LLaMA-2 on the MuMath dataset to derive the MuMath model. Our experiments indicate that our MuMath-70B model achieves new state-of-the-art performance among tool-free methods---achieving 84.5% on GSM8K (an increase of 2.2% compared to the previous best open-source LLM) and 32.2% on MATH (a rise by 5.6% compared to the prior best open-source LLM).
We release the MuMath dataset along with its corresponding models and code for public use.

Paper Type: long

Research Area: Machine Learning for NLP

Contribution Types: Data resources

Languages Studied: English