Anyone else find that LLMs have more trouble with declarative (as opposed to imperative) code? Maybe lack of training data or more difficult to follow chain-of-thought?
Which LLMs and languages did you try?
Please Login to reply.
I noticed this on o3 with Swift UI files. Have you had better luck with others?
I've mostly tried Claude Sonnet, haven't noticed a stark difference between imperative and declarative languages, at least it does well on SQL. I tried some functional programming, was quite ok
