logoalt Hacker News

mewpmewp210/11/20241 replyview on HN

I don't understand how it relates to temperature? Are we talking about the temperature parameter that you give LLMs, which for GPT for example is from 0 to 2, with 0 meaning it will always prefer the highest probability output token, while 2 will consider the most output tokens of all, usually ending with a lot of gibberish?

E.g. if I write "I have a cat and a "

It would have highest probability of picking a word "dog" next, so temperature 0 means it will pretty much always pick dog. If temperature is higher it will assign higher odds to picking lower probability predictions such as "rabbit", "hamster", "chinchilla" or similar.

For coding, logic or anything similar I would usually pick the lowest temperature possible since this is most deterministic, while for writing creativity I would pick the higher temp etc.


Replies

ActorNightly10/12/2024

Im saying temperature is a hack to make the models actually produce real answers.

show 1 reply