GuideBench: Benchmarking Domain-Oriented Guideline Following for LLM Agents

TL;DR

GuideBench is a new benchmark for evaluating large language models' ability to follow domain-specific guidelines, especially considering rule diversity, updates, and human alignment, addressing a key gap in current assessment tools.

Contribution

The paper introduces GuideBench, the first comprehensive benchmark for assessing LLMs' adherence to domain-oriented guidelines and their robustness to rule changes.

Findings

LLMs show significant room for improvement in guideline following.

Models vary widely in robustness to rule updates.

Alignment with human preferences is inconsistent across models.

Abstract

Large language models (LLMs) have been widely deployed as autonomous agents capable of following user instructions and making decisions in real-world applications. Previous studies have made notable progress in benchmarking the instruction following capabilities of LLMs in general domains, with a primary focus on their inherent commonsense knowledge. Recently, LLMs have been increasingly deployed as domain-oriented agents, which rely on domain-oriented guidelines that may conflict with their commonsense knowledge. These guidelines exhibit two key characteristics: they consist of a wide range of domain-oriented rules and are subject to frequent updates. Despite these challenges, the absence of comprehensive benchmarks for evaluating the domain-oriented guideline following capabilities of LLMs presents a significant obstacle to their effective assessment and further development. In this paper, we introduce GuideBench, a comprehensive benchmark designed to evaluate guideline following performance of LLMs. GuideBench evaluates LLMs on three critical aspects: (i) adherence to diverse rules, (ii) robustness to rule updates, and (iii) alignment with human preferences. Experimental results on a range of LLMs indicate substantial opportunities for improving their ability to follow domain-oriented guidelines.