Definition
Chain-of-thought prompting asks a model to think step-by-step before answering, dramatically improving performance on reasoning-heavy tasks. The trick works because the model is now using its own intermediate tokens as working memory, but the visible chain is not always the real chain.
Episodes covering this
Worth reading next
Papers we haven't done a deep dive on yet, but would recommend on this topic.
- Language Models are Few-Shot Learners
- Unfaithful Explanations in Chain-of-Thought Prompting
- Let's Verify Step by Step
- Faith and Fate: Limits of Transformers on Compositionality
- Let's Think Dot by Dot: Hidden Computation in Transformer Language Models
- To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning
- Language Agents Reduce the Frequency of Deliberative Reasoning
- Chain of Thought Empowers Transformers to be Expressive
- Training Large Language Models to Reason in a Continuous Latent Space
- Chain of Thought Prompting Elicits Reasoning in Large Language Models
- Let's Think Step by Step: Large Language Models are Zero-Shot Reasoners
- Scaling LLM Test-Time Compute Optimally Can be More Effective than Scaling Model Parameters
- Chain of Thought Empowers Transparent Reasoning for Large Language Models on Decision Tasks
- Measuring Faithfulness in Chain-of-Thought Reasoning
- Chain of Thought Empowers Transparent Reasoning of Language Models
- Calibration of Large Language Models Using Their Generations
- Self-Consistency Improves Chain of Thought Reasoning in Language Models