Are you sure about that? Chain of thought does not need to be semantically useful to improve LLM performance. https://arxiv.org/abs/2404.15758
replyIf you're misusing LLMs to solve TC^0 problems, which is what the paper is about, then... you also don't need the slop lavine. You can just inject a bunch of filler tokens yourself.
replystill doesn't mean all tokens are useful. it's the point of benchmarks
replyCare to share the benchmarks backing the claims in this repo?
reply