Chain-of-Thought: thinking before answering

Telling a model to "think step by step" changes everything on some tasks and nothing on others. You will measure where the difference lies.

Sources

• Wei, J. et al. 2022 · Chain-of-Thought Prompting Elicits Reasoning in Large Language Models · NeurIPS 2022
• Nye, M. et al. 2021 · Show Your Work: Scratchpads for Intermediate Computation with Language Models · arXiv:2112.00114
• Kojima, T. et al. 2022 · Large Language Models are Zero-Shot Reasoners · NeurIPS 2022
• Feng, G. et al. 2023 · Towards Revealing the Mystery behind Chain of Thought: A Theoretical Perspective · NeurIPS 2023