AutoPlan: Automatic Planning of Interactive Decision-Making Tasks With Large Language Models

TL;DR

AutoPlan introduces a novel method to enhance large language models in complex decision-making tasks by guiding them with optimized plans, achieving high success rates without relying on in-context demonstrations.

Contribution

AutoPlan is the first approach to optimize task-solving plans for LLMs through iterative experience, improving decision-making performance without costly demonstrations.

Findings

Achieves comparable success rates to human demonstration-based methods on ALFWorld.

Outperforms baseline methods by 8% on HotpotQA.

Operates without in-context demonstrations, reducing complexity.

Abstract

Recent large language models (LLMs) are promising for making decisions in grounded environments. However, LLMs frequently fail in complex decision-making tasks due to the misalignment between the pre-trained knowledge in LLMs and the actual rules in the environment. Existing methods require either costly gradient computation or lengthy in-context demonstrations. In this paper, we propose AutoPlan, an approach to guide LLM-based agents to accomplish interactive decision-making tasks. AutoPlan augments the LLM prompt with a task-solving plan and optimizes it through iterative experience collection and reflection. Our experiments show that AutoPlan, though using no in-context demonstrations, achieves success rates on par with the baselines using human-written demonstrations on ALFWorld and even outperforms them by 8% on HotpotQA. The code is available at https://github.com/owaski/AutoPlan.