If It Ain’t Broke, Don’t Fix It: the Unintended Consequences of Large Language Model Code Repairs

• Author(s) / Creator(s)

  • Xiaofei Yu
  • Thibaud Lutellier

• Nowadays, we heavily rely on ChatGPT to generate content, including writing code. But have you ever thought about the scenario where you input the correct content while GPT outputs a bug? This project aims to explore the unintended consequences of code repairs made by large language models. By comparing the original correct code with the code generated by the language model, we evaluate whether these repairs are accurate.

• Date created

  2024-08-01

• Subjects / Keywords

  • LLM
  • chatGPT
  • Code
  • Bug

• Type of Item

  Image

• DOI

  https://doi.org/10.7939/r3-218h-wg78

• License

  Attribution-NonCommercial 4.0 International