Temperature Scaling and Beam Search Text Generation in LLMs, for the ML-Adjacent | Towards Data Science

What “temperature” is, how it works, its relationship to the beam search heuristic, and where LLM output generation can still fail