Skip to main content
All terms
Foundations

Probability

A core mathematical idea behind how models score and sample possible outputs.

Definition

Probability is a core mathematical idea throughout machine learning. Models typically output a probability distribution over possible labels or next tokens, and training objectives like cross-entropy measure how well those predicted probabilities match reality. Sampling strategies such as temperature, top-p, and top-k all operate on these distributions. Calibration describes how well stated probabilities match observed frequencies.