PlanBench: An Extensible Benchmark for Evaluating Large Language Models on Planning and Reasoning about Change

TL;DR

PlanBench is a new, extensible benchmark suite designed to systematically evaluate large language models' planning and reasoning abilities across diverse domains, revealing current models' limitations in core planning tasks.

Contribution

The paper introduces PlanBench, a comprehensive benchmark inspired by automated planning domains, to assess LLMs' innate planning and reasoning skills beyond common sense tasks.

Findings

LLMs perform poorly on critical planning capabilities.

Current SOTA models show significant room for improvement in planning tasks.

PlanBench serves as a progress marker for LLM planning and reasoning abilities.

Abstract

Generating plans of action, and reasoning about change have long been considered a core competence of intelligent agents. It is thus no surprise that evaluating the planning and reasoning capabilities of large language models (LLMs) has become a hot topic of research. Most claims about LLM planning capabilities are however based on common sense tasks-where it becomes hard to tell whether LLMs are planning or merely retrieving from their vast world knowledge. There is a strong need for systematic and extensible planning benchmarks with sufficient diversity to evaluate whether LLMs have innate planning capabilities. Motivated by this, we propose PlanBench, an extensible benchmark suite based on the kinds of domains used in the automated planning community, especially in the International Planning Competition, to test the capabilities of LLMs in planning or reasoning about actions and change. PlanBench provides sufficient diversity in both the task domains and the specific planning capabilities. Our studies also show that on many critical capabilities-including plan generation-LLM performance falls quite short, even with the SOTA models. PlanBench can thus function as a useful marker of progress of LLMs in planning and reasoning.