Knowledge Distillation
Knowledge distillation trains a small student model to copy useful signals from a larger teacher model. Training often uses both teacher scores and known correct labels.
Another example
An edge-ML engineer trains a student with teacher scores and labels from 2 million sentences. The student keeps the measured intent quality and meets the phone speed limit. File compression or quantization alone is not distillation.
Main parts
- 01teacher model
- 02student model
- 03distillation objective
Use it when
Use this term when measured behavior is taught from a teacher model to a smaller student model.
Do not use it when
Do not assume the student gets every skill or safety feature from the teacher.
Name used in code
teacher logits → student loss Before you ship
Check the model and prompt versions, source data, versioned evaluation set, and measures. Verify links to evidence, tool permissions, privacy, common failures, decline-to-answer and fallback behavior, human review, monitoring, cost, speed, and rollback.
Outcome: Use or evaluate Knowledge Distillation to make AI behavior measurable and tied to evidence. User context: An edge-ML engineer trains a student with teacher scores and labels from 2 million sentences. The student keeps the measured intent quality and meets the phone speed limit. File compression or quantization alone is not distillation. AI method or concept: Knowledge Distillation. Why it fits: Use this term when measured behavior is taught from a teacher model to a smaller student model. Do not use it when: Do not assume the student gets every skill or safety feature from the teacher. AI requirements: Define the input and source evidence. Set model and tool permissions. Use versioned evaluation data and measures. Define failure, decline-to-answer, privacy, speed, and cost limits. Operational safeguards: Keep a clear trace and hide sensitive log data. Show users a safe fallback. Mark steps that need human review. Define how to roll back the model or prompt. Acceptance criteria: Record baseline and target measures on a versioned evaluation set. Test edge cases and hostile inputs. Verify fallback, monitoring, permissions, and rollback. Evidence and limits (evidence boundary): No single official standard controls the whole term. Its meaning may change across teams, platforms, or frameworks. Unknowns to confirm: Target task, model and version, evaluation owner, source data, risk limit, tool permissions, and production fallback.
Check this request
Review the current use of Knowledge Distillation. Definition: Knowledge distillation trains a small student model to copy useful signals from a larger teacher model. Training often uses both teacher scores and known correct labels. Release checks: Check the model and prompt versions, source data, versioned evaluation set, and measures. Verify links to evidence, tool permissions, privacy, common failures, decline-to-answer and fallback behavior, human review, monitoring, cost, speed, and rollback. Before changing code, report the evidence you found, gaps, severity, and the smallest safe fix.