alt Hacker NewsI feel like the LLMs used them "wrong". They use them for hard breaks for disconnected thoughts rather than one thought that disconnects.
I feel like the LLMs used them "wrong". They use them for hard breaks for disconnected thoughts rather than one thought that disconnects.
I think they just used them too much. I suspect that as they naturally form a thematic break point in writing, their probability was unnaturally high at the end of phrases and this led to overuse.
I don't ever recall seeing an LLM using them incorrectly, as opposed to simply at high frequency but correctly (although I'm sure some low param or early models must have done so at some point).