arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2402.16352cs.CLcs.AI
收藏

MathGenie:通过问题回译生成合成数据以增强大语言模型的数学推理能力

MathGenie: Generating Synthetic Data with Question Back-translation for Enhancing Mathematical Reasoning of LLMs

The Chinese University of Hong Kong(香港中文大学);Shanghai Artificial Intelligence Laboratory(上海人工智能实验室)

Zimu Lu, Aojun Zhou, Houxing Ren, Ke Wang, Weikang Shi, Junting Pan, Mingjie Zhan, Hongsheng Li

AI总结:

针对开源大模型数学推理弱于闭源模型的问题,提出MathGenie方法通过解答增强与问题回译生成合成数据,训练的MathGenieLM系列模型在多个数学推理基准上达到开源最优。

摘要:

大语言模型(LLMs)在数学推理领域展现出巨大潜力,但现有开源模型与GPT-4等闭源模型之间仍存在性能差距。本文提出MathGenie,这是一种从小规模问题-解答数据集(记为种子数据)生成多样且可靠数学题的新方法。我们对种子数据的真实解答进行增强,训练回译模型将增强后的解答回译为新问题,随后为新问题生成整合代码的解答。为确保整合代码解答的正确性,我们采用基于推理过程的策略进行解答验证。我们在新构建的数据集上训练了从7B到70B规模的多种预训练模型,以验证所提出增强技术的有效性,由此得到的模型系列名为MathGenieLM。这些模型在五个代表性数学推理数据集上始终优于此前的开源模型,达到了当前最优性能。其中,MathGenieLM-InternLM2在GSM8K上准确率达87.7%,在MATH上达55.7%,在开源语言模型中取得了最佳综合得分。

英文摘要:

Large language models (LLMs) have exhibited great potential in mathematical reasoning. However, there remains a performance gap in this area between existing open-source models and closed-source models such as GPT-4. In this paper, we introduce MathGenie, a novel method for generating diverse and reliable math problems from a small-scale problem-solution dataset (denoted as seed data). We augment the ground-truth solutions of our seed data and train a back-translation model to translate the augmented solutions back into new questions. Subsequently, we generate code-integrated solutions for the new questions. To ensure the correctness of the code-integrated solutions, we employ rationale-based strategy for solution verification. Various pretrained models, ranging from 7B to 70B, are trained on the newly curated data to test the effectiveness of the proposed augmentation technique, resulting in a family of models known as MathGenieLM. These models consistently outperform previous open-source models across five representative mathematical reasoning datasets, achieving state-of-the-art performance. In particular, MathGenieLM-InternLM2 achieves an accuracy of 87.7% on GSM8K and 55.7% on MATH, securing the best overall score among open-source language models.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑