Achieving >97% on GSM8K: Deeply Understanding the Problems Makes LLMs Better Solvers for Math Word Problems

TL;DR

This paper introduces DUP, a method that enhances LLMs' ability to solve math word problems by promoting deep understanding, leading to state-of-the-art accuracy on GSM8K and other benchmarks.

Contribution

The paper presents DUP, a novel approach focusing on reducing semantic misunderstanding errors to significantly improve LLMs' math reasoning performance.

Findings

DUP achieves 97.1% accuracy on GSM8K in zero-shot setting.

DUP outperforms existing methods across 10 diverse reasoning benchmarks.

Deep understanding of problems is key to improving LLM reasoning.

Abstract

Chain-of-Thought (CoT) prompting has enhanced the performance of Large Language Models (LLMs) across various reasoning tasks. However, CoT still falls short in dealing with complex math word problems, as it usually suffers from three pitfalls: semantic misunderstanding errors, calculation errors, and step-missing errors. Prior studies involve addressing the calculation errors and step-missing errors, but neglect the semantic misunderstanding errors, which is the major factor limiting the reasoning performance of LLMs. To this end, we propose a simple-yet-effective method, namely Deeply Understanding the Problems (DUP), to improve the LLMs' math problem-solving ability by addressing semantic misunderstanding errors. The core of our method is to encourage the LLMs to deeply understand the problems and extract the key problem-solving information used for better reasoning. Extensive experiments on 10 diverse reasoning benchmarks show that our DUP method consistently outperforms the other counterparts by a large margin. More encouragingly, DUP achieves a new SOTA result on the GSM8K benchmark, with an accuracy of 97.1% under the zero-shot setting.