Dr Manabu Sakamoto (he/him) on Nostr: LLM outputs should come with a metric of how good the response is, like probability ...
LLM outputs should come with a metric of how good the response is, like probability of a correct statement. Or even the individual probabilities of each word following another word.
Published at
2024-05-21 11:45:07Event JSON
{
"id": "3d4d3c448541a590c37ff5b86e63b01da764270153b8b3bdc77a83af4fd75603",
"pubkey": "53098d56dfdb2230c145d067df3a2e108553e9c9ba57a24c085e00f546f99f57",
"created_at": 1716291907,
"kind": 1,
"tags": [
[
"proxy",
"https://ecoevo.social/users/drmambobob/statuses/112478906450201387",
"activitypub"
]
],
"content": "LLM outputs should come with a metric of how good the response is, like probability of a correct statement. Or even the individual probabilities of each word following another word.",
"sig": "8a631361d8f8111f518b5013deab082ca7327e9b4aacb092dc9bd302937c6ade61d9f360fce5dea59f8ef873c0bc376a6884927d07a647719835f64b55cc476b"
}