English

PanGu-Coder2: Boosting Large Language Models for Code with Ranking Feedback

Computation and Language 2023-07-28 v1 Artificial Intelligence Machine Learning Programming Languages Software Engineering

Abstract

Large Language Models for Code (Code LLM) are flourishing. New and powerful models are released on a weekly basis, demonstrating remarkable performance on the code generation task. Various approaches have been proposed to boost the code generation performance of pre-trained Code LLMs, such as supervised fine-tuning, instruction tuning, reinforcement learning, etc. In this paper, we propose a novel RRTF (Rank Responses to align Test&Teacher Feedback) framework, which can effectively and efficiently boost pre-trained large language models for code generation. Under this framework, we present PanGu-Coder2, which achieves 62.20% pass@1 on the OpenAI HumanEval benchmark. Furthermore, through an extensive evaluation on CoderEval and LeetCode benchmarks, we show that PanGu-Coder2 consistently outperforms all previous Code LLMs.

Keywords

Cite

@article{arxiv.2307.14936,
  title  = {PanGu-Coder2: Boosting Large Language Models for Code with Ranking Feedback},
  author = {Bo Shen and Jiaxin Zhang and Taihong Chen and Daoguang Zan and Bing Geng and An Fu and Muhan Zeng and Ailun Yu and Jichuan Ji and Jingyang Zhao and Yuenan Guo and Qianxiang Wang},
  journal= {arXiv preprint arXiv:2307.14936},
  year   = {2023}
}

Comments

Preprint

R2 v1 2026-06-28T11:41:57.835Z