CoDAT: Code maintenance, synthesis, and verification via LLM-based documentation

Keywords: Large language models, code generation, code verification, code documentation

TL;DR: Presents a tool, CoDAT, for documenting code and using LLM's to check that the code implements the documentation correctly.

Abstract:

We present a a prototype tool, CoDAT,
in which code can be linked to corresponding documentation, and the query
"does the code correctly implemented the associated documentation"? can be posed to several large language models. The answers to the are either ``yes, the code does correctly implement its documentation", or "no", the code does not correctly implement its documentation". These are displayed graphically, which enables a "birds eye view" in which the results of many queries can be viewed simultaneously. We show how CoDAT can be used to document, verify, and generate code.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public.

Paper Type: Full-length papers (i.e. case studies, theoretical, applied research papers). 8 pages

Reroute: false

Submission Number: 22