Reproduce “Retrieval-Augmented Code Generation: A Survey with Focus on Repository-Level Approaches” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yicheng Tao, Yuante Li, Yao Qin, Yepang Liu
This survey systematically reviews Retrieval-Augmented Code Generation (RACG) with a focus on repository-level applications, introducing a unified analytical framework and categorizing existing methods across retrieval strategies, training optimizations, and autonomous agent architectures. It provides a comprehensive overview of the field and identifies key challenges and future research directions for AI-powered software engineering.
Key insights
- Retrieval-Augmented Code Generation (RACG) is characterized as a dynamic, coupled process where retrieval and generation are interdependent, moving beyond static pipelines to incorporate iterative refinement.
- RACG strategies are broadly categorized into non-graph-based methods (treating repositories as text for similarity matching) and graph-based methods (leveraging explicit structural code information), each with distinct content construction and optimization techniques.
- There is an observed shift towards autonomous agent architectures for RLCG, employing iterative feedback, tool augmentation, and multi-agent coordination for dynamic problem-solving and solution refinement.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.