Tailoring with Targeted Precision: Edit-Based Agents for Open-Domain Procedure Customization

TL;DR

This paper evaluates multi-agent LLM architectures for customizing procedural instructions, finding that a sequential editing and verification approach outperforms end-to-end models in procedure adaptation tasks.

Contribution

It introduces a new evaluation set and demonstrates that a simple multi-agent editing architecture significantly improves customization performance over end-to-end LLMs.

Findings

Sequential multi-agent architecture outperforms end-to-end LLMs by 10.5%

Two-agent system effectively edits and verifies procedures

Multi-agent editing architectures are promising for various customization tasks

Abstract

How-to procedures, such as how to plant a garden, are now used by millions of users, but sometimes need customizing to meet a user's specific needs, e.g., planting a garden without pesticides. Our goal is to measure and improve an LLM's ability to perform such customization. Our approach is to test several simple multi-LLM-agent architectures for customization, as well as an end-to-end LLM, using a new evaluation set, called CustomPlans, of over 200 WikiHow procedures each with a customization need. We find that a simple architecture with two LLM agents used sequentially performs best, one that edits a generic how-to procedure and one that verifies its executability, significantly outperforming (10.5% absolute) an end-to-end prompted LLM. This suggests that LLMs can be configured reasonably effectively for procedure customization. This also suggests that multi-agent editing architectures may be worth exploring further for other customization applications (e.g. coding, creative writing) in the future.