Open Internet by MindsNet

Improving AI-generated code reliability

AI-generated code can be unreliable due to small wording changes. Existing models across Java, C++, and JavaScript show limitations in robustness. Larger models do not consistently perform better. This issue affects the dependability of automated code generation.

Computing & Technology, Computer Science, Artificial Intelligence