Llama 2 as a Next-Generation Partial Order Programming Tool: Performance Benchmarks and Implications for Autonomous Systems

Examining the planning and reasoning abilities of large language models (LLMs) has become a notable area of research due to their crucial role in the functionality of intelligent agents. Consequently, the assessment of planning and reasoning capabilities in LLMs has obtained significant attention in the research field. While most evaluations of LLM planning abilities focus on everyday tasks, distinguishing between genuine planning and knowledge retrieval from their extensive databases poses a challenge. The aim of this study is to determine the extent to which LLMs can autonomously solve planning problems and to understand how LLMs can guide AI planners and whether their planning performance can be enhanced to rival that of dedicated plan-solving algorithms. This study aims to explore the potential of leveraging LLMs as Partial Order Planning tools and investigate the efficacy of fine-tuning techniques in enhancing model performance.