This repository curates essential research papers on the planning capabilities of Large Language Models (LLMs). It serves as a valuable resource for researchers and practitioners in AI, NLP, and robotics seeking to understand and leverage LLMs for complex task execution and decision-making.

How It Works

The collection highlights papers that explore LLMs as zero-shot planners, investigate prompting strategies like "Least-to-Most" and "Plan-and-Solve," and examine LLMs' ability to generate actionable knowledge for embodied agents. Approaches include grounding LLM planning in real-world tasks, integrating planning with world models, and developing benchmarks for evaluating planning proficiency.

Quick Start & Requirements

• Access: The repository provides a curated list of papers with links to their abstracts and, where available, code repositories.
• Requirements: Access to academic paper repositories (e.g., arXiv, conference proceedings) and potentially code execution environments for linked projects.

Highlighted Details

• Comprehensive list of 30+ papers published between ICML 2022 and Preprint 2023.10.
• Covers diverse planning paradigms: zero-shot, few-shot, grounded, multimodal, and hierarchical.
• Includes papers focusing on specific applications like embodied agents, web agents, and dialogue systems.
• Features foundational work on LLM reasoning, world models, and benchmark development for planning.

Maintenance & Community

• The list was last updated on August 12, 2023.
• No specific community channels or contributor information are detailed in the README.

Licensing & Compatibility

• The repository itself does not host code or data, thus no specific license is applied to the curated list. Individual papers retain their original licensing.

Limitations & Caveats

• This is a curated list of papers, not a runnable software project. The "quick start" is for accessing research, not for deploying an LLM planner.
• The list may not be exhaustive and is subject to the last update date.