VERSE: Verification-based Self-Play for Code Instructions

Instruction-tuned Code Large Language Models (Code LLMs) have excelled in diverse code-related tasks, such as program synthesis, automatic program repair, and code explanation. To collect training datasets for instruction-tuning, a popular method involves having models autonomously generate instructions and corresponding responses. However, the direct generation of responses does not ensure functional correctness, a crucial requirement for generating responses to code instructions. To overcome this, we present Verification-Based Self-Play (VERSE), aiming to enhance model proficiency in generating correct responses. VERSE establishes a robust verification framework that covers various code instructions. Employing VERSE, Code LLMs engage in self-play to generate instructions and corresponding verifications. They evaluate execution results and self-consistency as verification outcomes, using them as scores to rank generated data for self-training. Experiments show that VERSE improves multiple base Code LLMs (average 7.6%) across various languages and tasks on many benchmarks, affirming its effectiveness.

Authors

• Hao Jiang Xiaomi Inc., Beijing, China
• Qi Liu
• Rui Li State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China
• Yuze Zhao Alibaba Group
• Yixiao Ma State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China
• Shengyu Ye State Key Laboratory of Cognitive Intelligence, University of Science and Technology of China
• Junyu Lu Institute of Artificial Intelligence, Hefei Comprehensive National Science Center
• Yu Su Institute of Artificial Intelligence, Hefei Comprehensive National Science Center School of Computer Science and Artificial Intelligence, Hefei Normal University

Keywords

No keywords are indexed for this paper.