Infinite-Instruct: Synthesizing Scaling Code instruction Data with Bidirectional Synthesis and Static Verification

TL;DR

Infinite-Instruct is a scalable framework that synthesizes high-quality, logically consistent code instruction data using bidirectional methods and static verification, significantly improving code generation performance of large language models.

Contribution

The paper introduces a novel automated framework combining reverse and backfeeding construction with static code analysis to generate diverse, high-quality code instruction datasets for LLM training.

Findings

Achieved up to 36.95% performance improvement on code benchmarks.

Generated datasets enable comparable performance with less fine-tuning data.

Open-sourced datasets facilitate further research in code instruction synthesis.

Abstract

Traditional code instruction data synthesis methods suffer from limited diversity and poor logic. We introduce Infinite-Instruct, an automated framework for synthesizing high-quality question-answer pairs, designed to enhance the code generation capabilities of large language models (LLMs). The framework focuses on improving the internal logic of synthesized problems and the quality of synthesized code. First, "Reverse Construction" transforms code snippets into diverse programming problems. Then, through "Backfeeding Construction," keywords in programming problems are structured into a knowledge graph to reconstruct them into programming problems with stronger internal logic. Finally, a cross-lingual static code analysis pipeline filters invalid samples to ensure data quality. Experiments show that on mainstream code generation benchmarks, our fine-tuned models achieve an average performance improvement of 21.70% on 7B-parameter models and 36.95% on 32B-parameter models. Using less than one-tenth of the instruction fine-tuning data, we achieved performance comparable to the Qwen-2.5-Coder-Instruct. Infinite-Instruct provides a scalable solution for LLM training in programming. We open-source the datasets used in the experiments, including both unfiltered versions and filtered versions via static analysis. The data are available at https://github.com/xingwenjing417/Infinite-Instruct-dataset