Generative AI and Post-Training
Connect generative objectives, flow matching, multimodal alignment, preference learning, GRPO, compression, and interpretability to measurable model behavior.
Begin chapterPrerequisites
- 01Generative Models
GANs, VAEs, and Diffusion Probabilistic Models (DDPM/DDIM).
- 02Flow Matching & DiT Video
Rectified Flow matching, optimal transport vector fields, and Patchified Diffusion Transformers (Sora/Flux).
- 03Multimodal AI & CLIP
Contrastive vision-text embeddings, InfoNCE loss, and joint cross-modal representation spaces.
- 04RLHF
Reinforcement Learning from Human Feedback.
- 05Frontier Reasoning & GRPO
DeepSeek-R1 style Group Relative Policy Optimization, rule-based RL, and test-time compute.
- 06Quantization & LoRA
Low-Rank Adaptation (ΔW = B·A), rank selection, FP8/INT4 uniform quantization, and scale-zero point math.
- 07Mixture of Experts (MoE)
Sparse routing mechanisms, gating networks, Top-K dispatch, and auxiliary load balancing.
- 08Mechanistic Interpretability
Sparse Autoencoders (SAEs), monosemantic feature dictionaries, circuit analysis, and activation steering.