Chain-of-Thought Reasoning
A prompting technique where LLMs solve problems by generating intermediate reasoning steps.
Overview
Chain-of-thought (CoT) reasoning is a technique where large language models are prompted to generate intermediate reasoning steps before arriving at a final answer. Instead of jumping directly to a conclusion, the model explains its thinking step by step, dramatically improving performance on math, logic, and complex reasoning tasks.
Key Details
CoT can be elicited through few-shot examples that demonstrate step-by-step reasoning, or through zero-shot prompting with phrases like 'Let's think step by step.' Variants include self-consistency (sampling multiple reasoning chains and taking the majority answer), tree of thoughts (exploring branching reasoning paths), and process reward models that evaluate each reasoning step. CoT is now standard practice for complex tasks with LLMs.