Project-specific code summarization with in-context learning

Shangbo Yun, Shuhuai Lin, Xiaodong Gu, Beijun Shen

Abstract: Highlights•The first to tackle the few-shot problem of code summarization for a specific project.•Explores critical factors to high-quality prompts in large language models for project-specific code summarization.•Automatically generates in-context learning examples for a specific project by using large language models.•Designs a prompt selector by incorporating P-tuning into CodeBERT.•Demonstrates superior performance in project-specific code summarization task compared to state-of-the-art approaches.