Following multiple instructions is a crucial ability for large language models (LLMs). Evaluating this ability comes with significant challenges: (i) limited coherence between multiple instructions, (ii) positional bias where the order of instructions affects model performance, and (iii) a lack of objectively verifiable tasks. To address these issues, we introduce a benchmark designed to evaluate models' abilities to follow multiple instructions through sequential instruction following (SIFo) tasks. In SIFo, the successful completion of multiple instructions is verifiable by examining only the final instruction. Our benchmark evaluates instruction following using four tasks (text modification, question answering, mathematics, and security rules), each assessing different aspects of sequential instruction following. Our evaluation of popular LLMs, both closed-source and open-source, shows that more recent and larger models significantly outperform their older and smaller counterparts on the SIFo tasks, validating the benchmark's effectiveness. All models struggle with following sequences of instructions, hinting at an important lack of robustness of today's language models.
@article{arxiv.2406.19999,
title = {The SIFo Benchmark: Investigating the Sequential Instruction Following Ability of Large Language Models},
author = {Xinyi Chen and Baohao Liao and Jirui Qi and Panagiotis Eustratiadis and Christof Monz and Arianna Bisazza and Maarten de Rijke},
journal= {arXiv preprint arXiv:2406.19999},
year = {2025}
}