Atlas · GenAI 2026

Knowledge Distillation

Knowledge Distillation

conceptPeak: 2020Model CompressionAI consensus: 2/3

Prerequisites

  • Distillation trains a student network to mimic a teacher network — both are neural networks requiring DL understanding

  • Evaluating distillation quality requires comparing student vs. teacher on meaningful metrics

Recommended reference

Hinton et al. (2015) 'Distilling the Knowledge in a Neural Network' — the foundational paper. Updated: Gu et al. (2024) 'MiniLLM: Knowledge Distillation of LLMs'

Notes from AI deep research

Anthropic Opus

Hinton (2015). Teacher → student. W 2026: distillacja frontier do szybkich lokalnych 8B

Google Deep Think

Zdolności modelu 100B w 8B [G#67]

Related skills

  • → is subcategory of: Model Fine-Tuning(3/3)
  • → is part of: Edge AI(2/3)
  • → is subcategory of: LLM Fine-Tuning(1/3)