Reasoning Techniques: Chain of Thought, Self-Consistency, and Tree of Thoughts
~14 min read
The book's 3 named prompting techniques for reasoning — CoT (reason step by step), Self-Consistency (vote across multiple reasoning paths), and Tree of Thoughts (search across branching reasoning paths) — each building on the last.
Reasoning Techniques: Chain of Thought, Self-Consistency, and Tree of Thoughts is a Pro topic
Sign in, then upgrade to Pro or Power to unlock this topic and the full AI Engineering curriculum.
Key points
- •Chain of Thought (CoT): nudge the model to reason step by step instead of jumping to the answer — the simplest, most widely used technique
- •Self-Consistency: since CoT isn't always consistent across runs, generate multiple reasoning paths and take a majority vote on the FINAL ANSWER — but it doesn't evaluate the reasoning quality itself
- •Tree of Thoughts (ToT): unlike Self-Consistency (varies the final answer), ToT varies the STEPS at each point, forming a branching tree evaluated by a separate process — more compute-intensive but stronger
- •The three form a progression: CoT (one path) -> Self-Consistency (many independent full paths, vote) -> ToT (one branching, evaluated search tree)
- •All three still rely on free-form thinking, which breaks down on long, rule-heavy tasks — the gap that ARQ (covered in this curriculum's separate prompt-reasoning topic) addresses with structured queries instead