Code-Survey: An LLM-Driven Methodology for Analyzing Large-Scale Codebases
arxiv.orgI think this paper is well written and the data is solid.Provides a new and valuable perspective. This paper, introduces an innovative approach for exploring and understanding large, evolving software systems using Large Language Models (LLMs). The key contributions include:
- Automating the transformation of unstructured codebase data into structured, analyzable datasets. - Uncovering valuable insights into the evolution of the Linux kernel's eBPF subsystem, validated by domain experts. - Proposing a versatile methodology applicable to other large-scale software projects.
By leveraging LLMs, the paper demonstrates how to address challenges in understanding software evolution, feature interdependencies, and design decisions efficiently.