# Chain-of-Thought (CoT) prompting Chain-of-Thought (CoT) is an effective prompting technique. CoT tell [[Large Language Models (LLMs)]] to “use step by step reasoning”, the way a human would, to solve a problem. With CoT, LLMs do their best impressions of deduction. And because LLMs are good at impressions, they often guess correct lines of deduction. This makes the conclusions more likely to be correct. A written chain of thought has a useful side effect: you can read it, and check why the model reached its answer. [[AI Reasoning Models]] now do this natively, and research on [[Latent Reasoning]] tries to move it into hidden vectors, which saves tokens but removes that view. ## References - ## Related - [[AI Reasoning Models]] - [[Latent Reasoning]] - [[Hybrid Reasoning Agents]]