Enregistré dans:
Détails bibliographiques
Auteurs principaux: Wang, Jiankang, Xu, Jianjun, Wang, Xiaorui, Wang, Yuxin, Xing, Mengting, Fang, Shancheng, Xie, Hongtao
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:https://arxiv.org/abs/2412.08864
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866912597670887424
author Wang, Jiankang
Xu, Jianjun
Wang, Xiaorui
Wang, Yuxin
Xing, Mengting
Fang, Shancheng
Xie, Hongtao
author_facet Wang, Jiankang
Xu, Jianjun
Wang, Xiaorui
Wang, Yuxin
Xing, Mengting
Fang, Shancheng
Xie, Hongtao
contents Large-scale, high-quality data is essential for advancing the reasoning capabilities of large language models (LLMs). As publicly available Internet data becomes increasingly scarce, synthetic data has emerged as a crucial research direction. However, existing data synthesis methods often suffer from limited scalability, insufficient sample diversity, and a tendency to overfit to seed data, which constrains their practical utility. In this paper, we present \textit{\textbf{GRIP}}, a \textbf{G}raph-based \textbf{R}easoning \textbf{I}nstruction \textbf{P}roducer that efficiently synthesizes high-quality and diverse reasoning instructions. \textit{GRIP} constructs a knowledge graph by extracting high-level concepts from seed data, and uniquely leverages both explicit and implicit relationships within the graph to drive large-scale and diverse instruction data synthesis, while employing open-source multi-model supervision to ensure data quality. We apply \textit{GRIP} to the critical and challenging domain of mathematical reasoning. Starting from a seed set of 7.5K math reasoning samples, we construct \textbf{GRIP-MATH}, a dataset containing 2.1 million synthesized question-answer pairs. Compared to similar synthetic data methods, \textit{GRIP} achieves greater scalability and diversity while also significantly reducing costs. On mathematical reasoning benchmarks, models trained with GRIP-MATH demonstrate substantial improvements over their base models and significantly outperform previous data synthesis methods.
format Preprint
id arxiv_https___arxiv_org_abs_2412_08864
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle GRIP: A Graph-Based Reasoning Instruction Producer
Wang, Jiankang
Xu, Jianjun
Wang, Xiaorui
Wang, Yuxin
Xing, Mengting
Fang, Shancheng
Xie, Hongtao
Computation and Language
Large-scale, high-quality data is essential for advancing the reasoning capabilities of large language models (LLMs). As publicly available Internet data becomes increasingly scarce, synthetic data has emerged as a crucial research direction. However, existing data synthesis methods often suffer from limited scalability, insufficient sample diversity, and a tendency to overfit to seed data, which constrains their practical utility. In this paper, we present \textit{\textbf{GRIP}}, a \textbf{G}raph-based \textbf{R}easoning \textbf{I}nstruction \textbf{P}roducer that efficiently synthesizes high-quality and diverse reasoning instructions. \textit{GRIP} constructs a knowledge graph by extracting high-level concepts from seed data, and uniquely leverages both explicit and implicit relationships within the graph to drive large-scale and diverse instruction data synthesis, while employing open-source multi-model supervision to ensure data quality. We apply \textit{GRIP} to the critical and challenging domain of mathematical reasoning. Starting from a seed set of 7.5K math reasoning samples, we construct \textbf{GRIP-MATH}, a dataset containing 2.1 million synthesized question-answer pairs. Compared to similar synthetic data methods, \textit{GRIP} achieves greater scalability and diversity while also significantly reducing costs. On mathematical reasoning benchmarks, models trained with GRIP-MATH demonstrate substantial improvements over their base models and significantly outperform previous data synthesis methods.
title GRIP: A Graph-Based Reasoning Instruction Producer
topic Computation and Language
url https://arxiv.org/abs/2412.08864