IHEval: Evaluating Language Models on Following the Instruction Hierarchy

TL;DR

This paper introduces IHEval, a comprehensive benchmark to evaluate how well language models follow instruction hierarchies, revealing significant challenges and performance drops when models face conflicting instructions.

Contribution

The paper presents IHEval, the first extensive benchmark for assessing language models' ability to follow instruction hierarchies and handle conflicts.

Findings

Models struggle with instruction priority conflicts.

Performance drops sharply on conflicting instructions.

Open-source models achieve only 48% accuracy on conflicts.

Abstract

The instruction hierarchy, which establishes a priority order from system messages to user messages, conversation history, and tool outputs, is essential for ensuring consistent and safe behavior in language models (LMs). Despite its importance, this topic receives limited attention, and there is a lack of comprehensive benchmarks for evaluating models' ability to follow the instruction hierarchy. We bridge this gap by introducing IHEval, a novel benchmark comprising 3,538 examples across nine tasks, covering cases where instructions in different priorities either align or conflict. Our evaluation of popular LMs highlights their struggle to recognize instruction priorities. All evaluated models experience a sharp performance decline when facing conflicting instructions, compared to their original instruction-following performance. Moreover, the most competitive open-source model only achieves 48% accuracy in resolving such conflicts. Our results underscore the need for targeted optimization in the future development of LMs.