Temperature (in Generation)

Appears in 1 paper · 4 tutorials

A hyperparameter controlling randomness in generation.

As used in Paper 12 — Language Models are Few-Shot Learners →

A hyperparameter controlling randomness in generation. Low temperature (~0.1) makes sampling deterministic and greedy (always pick the highest probability). High temperature (~2.0) makes the model sample more randomly, producing more diverse outputs.

As used in AI Agents →

A setting controlling randomness. Low = focused/consistent (good for agents); high = creative/varied.

As used in Prompt Engineering →

A setting that controls randomness. Low = focused and consistent; high = varied and creative.

As used in AI Production Engineering →

A setting controlling randomness/creativity. Low = focused and repeatable; high = varied and creative. (Mod 1)

As used in LLM Infrastructure →

A knob (≈0–2) for randomness: low = focused/repetitive, high = creative/varied.