What happens to an LLM's sampled output when you lower the temperature?
- It gets more deterministic
- It gets much more random
- It gets much longer
- It stops following the prompt
Lower temperature sharpens the token distribution, so the most likely tokens are chosen more often.
Play AI & ML ▸