Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

Abstract

We explore how generating a chain of thought — a series of intermediate reasoning steps — significantly improves the ability of large language models to perform complex reasoning. We show that chain-of-thought prompting substantially outperforms standard prompting on arithmetic, commonsense, and symbolic reasoning benchmarks, with improvements most dramatic in the largest models.

Key Findings

• 1Demonstrated that chain-of-thought prompting dramatically improves LLM reasoning
• 2Showed that providing step-by-step reasoning examples unlocks emergent capabilities
• 3Achieved state-of-the-art on GSM8K math benchmark with prompting alone
• 4Found that chain-of-thought is an emergent ability appearing primarily in large models
• 5Required no fine-tuning, only changes to the prompt format

Impact & Significance