Skywork-Math：大型语言模型中数学推理的数据扩展定律——故事仍在继续
Skywork-Math: Data Scaling Laws for Mathematical Reasoning in Large Language Models -- The Story Goes On
AI总结：
本文提出Skywork-Math模型系列，通过250万实例的Skywork-MathQA数据集和两阶段数据合成流程，证明数学推理的数据扩展定律未饱和，在MATH和GSM8K基准上分别达到51.2%和83.9%的准确率，超越GPT-4早期版本。
摘要：
在本文中，我们研究了可能增强大型语言模型（LLMs）数学推理能力的潜在因素。我们认为，现代LLMs中数学推理能力的数据扩展定律远未饱和，强调了模型质量如何随着数据量的增加而提高。为支持这一观点，我们引入了Skywork-Math模型系列，该系列使用我们提出的包含250万实例的Skywork-MathQA数据集，对常见的7B LLMs进行监督微调（SFT）。Skywork-Math 7B在竞赛级别的MATH基准上取得了51.2%的惊人准确率，在GSM8K基准上取得了83.9%的准确率，仅使用SFT数据，就在MATH上超越了GPT-4的早期版本。Skywork-Math模型的卓越性能归功于我们新颖的两阶段数据合成和模型SFT流程，其中包括三种不同的增强方法和一个多样化的种子问题集，确保了Skywork-MathQA数据集在不同难度级别上的数量和质量。最重要的是，我们提供了几个实用的要点，以增强LLMs在研究和工业应用中的数学推理能力。
英文摘要：
In this paper, we investigate the underlying factors that potentially enhance the mathematical reasoning capabilities of large language models (LLMs). We argue that the data scaling law for math reasoning capabilities in modern LLMs is far from being saturated, highlighting how the model's quality improves with increases in data quantity. To support this claim, we introduce the Skywork-Math model series, supervised fine-tuned (SFT) on common 7B LLMs using our proposed 2.5M-instance Skywork-MathQA dataset. Skywork-Math 7B has achieved impressive accuracies of 51.2% on the competition-level MATH benchmark and 83.9% on the GSM8K benchmark using only SFT data, outperforming an early version of GPT-4 on MATH. The superior performance of Skywork-Math models contributes to our novel two-stage data synthesis and model SFT pipelines, which include three different augmentation methods and a diverse seed problem set, ensuring both the quantity and quality of Skywork-MathQA dataset across varying difficulty levels. Most importantly, we provide several practical takeaways to enhance math reasoning abilities in LLMs for both research and industry applications.
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。
