[glenneroo]

I've read from several supposed AI prompt-masters that this actually reduces output quality. I can't speak to the validity of these claims though.

[SquareWheel]

Forcing shorter answers will definitely reduce their quality. Every token an LLM generates is like a little bit of extra thinking time. Sometimes it needs to work up to an answer. If you end a response too quickly, such as by demanding one-word answers, it's much more likely to produce hallucinations.

[profunctor]

Is this proven?