Why LLMs Think Longer?

Learn how modern language models use extra inference-time computation to improve reasoning, why those methods work best on verifiable tasks, and where visible chain-of-thought can mislead. This course is for technically curious learners with basic LLM familiarity.

4 sections ยท 11 lessons

Course outline

Compute Budget

  1. Slow Thinking for Models
  2. Inference Compute as a Resource

Reasoning Paths

  1. Chain-of-Thought as Scratchpad
  2. Branching and Revision

Training and Trust

  1. RL on Verifiable Tasks
  2. Tools as Reliable Substeps
  3. Faithfulness Under Pressure

Hidden and Scaled

  1. Thinking Beyond Words
  2. Latent Thoughts and EM
  3. Self-Taught Reasoning Loops
  4. When Thinking Longer Scales

Start learning with Wondering