Data/ML/AITable-stakes
Temperature / Sampling Parameters
Also written as LLM Temperature, Top-p, Sampling Parameters, Nucleus Sampling
Settings that control how predictable or varied a language model's output is. Low temperature makes it pick the most likely words, so answers are consistent; high temperature makes it more creative and more random.
Think of it like
A dial between 'always order your usual' and 'try something different on the menu every time'.
Junior or senior?
Basic knowledge for anyone who has called an LLM API.
Senior sounds like
Can say what they set it to for a specific feature and why, and knows it doesn't fix a hallucination problem.
Ask them
“What temperature did you use in production, and what made you pick it?”