Chapter 3 Complete
You have learned
Reasoning & Test-Time Compute
You now understand how modern LLMs solve complex problems:
Chain-of-Thought prompting, hidden reasoning, and how
test-time compute improves performance on difficult tasks.
Chain-of-Thought (CoT)
o1/o3 Hidden Reasoning
DeepSeek R1 & GRPO
Compute Allocation
Test-Time Scaling
Effort Parameter
Thinking Budget
Dual-Mode Models
Continue with Chapter 4
Optimizations & Memory
Learn the techniques that make LLMs fast and memory-efficient:
KV-Cache, RoPE, ALiBi, Sliding Window Attention, Ring Attention,
Paged Attention, and RAG pipelines.