Does this technique limit the LLM to correctness-preserving transforms?

alt Hacker News

amelius • yesterday at 10:09 PM • 0 replies • view on HN