Just a random guess: is it possible that the tasks where LLMs benefit from chain-of-thought are the same tasks where mild autism is an advantage for humans? Like, maybe autism makes it easier for humans to chain the thoughts, at the expense of something else?
Just a random guess: is it possible that the tasks where LLMs benefit from chain-of-thought are the same tasks where mild autism is an advantage for humans? Like, maybe autism makes it easier for humans to chain the thoughts, at the expense of something else?