LLM 2: Training & Alignment — all 15 problems

Fine-tuning, LoRA, reward modeling, and preference optimization.