Ask HN: Are new lines useless with LLMs?
To LLMs experts, I'm wondering if new lines (\n) are useless when prompting LLMs like GPT-4.
They are indeed very token-consuming, so knowing if it's possible to omit them is nice.
Of course I'm wondering about generation performances if omitting
1 comment
[ 2.1 ms ] story [ 13.9 ms ] thread