LLMs Can't Plan, But Can Help Planning in LLM-Modulo Frameworks

arxiv:2402.01817

Original source · Paper JSON

Reasoning and the "physics" of language models · 2024-02-02

Cites (2) · Cited by (0)

Complete cleaned Markdown · Source Markdown including bibliography

Read in parts

1. Part 1
2. Part 2
3. Part 3

Contents

• Position: LLMs Can’t Plan, But Can Help Planning in LLM-Modulo Frameworks
• Abstract
• Keywords:
• 1 Introduction
• 2 Planning-centered Limitations of LLMs
• 2.1 LLMs cannot generate executable plans in autonomous mode
• 2.2 LLMs cannot verify plans and thus cannot improve by self-critiquing
• 2.3 Analyzing Claims to the Contrary in the Literature
• 3 LLM-Modulo Framework for Robust Planning
• 3.1 Critics/Verifers
• 3.2 Backprompt (Meta) Controller
• 3.3 Specification Refinement & Critic/Model Acquisition
• 3.4 Summary of LLM Roles in LLM-Modulo
• 3.5 Can LLM-Modulo Frameworks Pay Their Way?
• 4 Two Case Studies of LLM-Modulo
• 5 Related Work
• 6 Conclusion
• Impact Statement
• Acknowledgments