Procedural Knowledge Improves Agentic LLM Workflows

TL;DR

This paper demonstrates that integrating hierarchical task networks (HTNs) as procedural knowledge significantly enhances large language models' performance on agentic tasks, often surpassing larger models without HTNs.

Contribution

The paper formalizes and evaluates an agentic LLM workflow using HTNs, showing substantial performance improvements and the potential of procedural knowledge in LLM applications.

Findings

Hand-coded HTNs greatly improve LLM performance on agentic tasks.

Using HTNs can enable smaller LLMs to outperform larger baseline models.

LLM-generated HTNs also enhance performance, though to a lesser extent.

Abstract

Large language models (LLMs) often struggle when performing agentic tasks without substantial tool support, prom-pt engineering, or fine tuning. Despite research showing that domain-dependent, procedural knowledge can dramatically increase planning efficiency, little work evaluates its potential for improving LLM performance on agentic tasks that may require implicit planning. We formalize, implement, and evaluate an agentic LLM workflow that leverages procedural knowledge in the form of a hierarchical task network (HTN). Empirical results of our implementation show that hand-coded HTNs can dramatically improve LLM performance on agentic tasks, and using HTNs can boost a 20b or 70b parameter LLM to outperform a much larger 120b parameter LLM baseline. Furthermore, LLM-created HTNs improve overall performance, though less so. The results suggest that leveraging expertise--from humans, documents, or LLMs--to curate procedural knowledge will become another important tool for improving LLM workflows.