All terms

Knowledge Distillation

Training a smaller student model to mimic a larger teacher — compressing capability into a cheaper model.

Machine Learning1 min read

Definition

Distillation transfers behavior from a teacher (large model) to a student (smaller model) by matching soft probabilities, hidden states, or generated outputs — not only hard labels.

It is widely used to deploy fast classifiers and increasingly to build smaller LLMs.

In simple terms

A master chef trains an apprentice — the apprentice learns shortcuts and taste judgments, not just cookbook recipes.

Where you see it

  • Mobile vision models distilled from large CNNs.
  • Small chat models trained on teacher demonstrations.

How it works

  1. 1.Run the teacher

    Collect soft labels or traces.

  2. 2.Train the student

    Match teacher outputs plus task loss.

  3. 3.Deploy student

    Cheaper latency and memory.

Why it matters

  • Distillation bridges research-scale models and production constraints.

Often confused

  • The student always equals the teacher.

    Students approximate teachers — expect some capability gap.