Learning Center

Foundations

• 01What Is Knowledge Distillation? A Complete Guide12 min
• 02Teacher and Student Models: How Knowledge Transfer Actually Works9 min
• 03Soft Targets and Dark Knowledge: Why Temperature Scaling Matters14 min
• 04The Three Types of Knowledge Distillation, Compared10 min

Distillation in practice

• 05Distillation vs Fine-Tuning: Which Do You Actually Need?11 min
• 06Distillation vs RAG vs Fine-Tuning: Choosing an Approach12 min
• 07Black-Box vs White-Box Distillation: What Changes Without Logits11 min
• 08How Chain-of-Thought Reasoning Is Distilled Into Smaller Models13 min
• 09Synthetic Data vs Distillation: Where Is the Line?15 min
• 10How Much Does It Cost to Distill a Model?9 min
• 11LoRA vs Distillation: Two Different Problems8 min

Model compression

• 12What Is Quantization? int8, int4, GPTQ and AWQ Explained12 min
• 13Model Pruning Explained: Magnitude, Structured and the Lottery Ticket Hypothesis11 min
• 14What Is Speculative Decoding, and Why Does It Make Inference Faster?9 min
• 15Mixture of Experts Explained: Why MoE Models Are Cheaper to Run13 min

Detection, defence and provenance

• 16Model Extraction Attacks: The Research Behind the Headlines14 min
• 17How AI Watermarking Works, and How It Breaks16 min
• 18How to Tell If a Model Was Distilled From Another12 min
• 19Can You Prove a Model Was Copied? The Attribution Problem13 min

Law and policy

• 20Is AI Training Fair Use? What Bartz v. Anthropic Actually Decided14 min
• 21Open Weights vs Open Source: Why the Distinction Matters11 min
• 22Terms of Service and Model Outputs: What the Contracts Actually Say11 min
• 23Are Model Weights Intellectual Property?12 min
• 24Export Controls and Sovereign AI14 min
• 25What Is a Frontier Model?7 min
• 26Case Study: The Distillation Accusations of 2025 to 202622 min