XIFBench: Evaluating Large Language Models on Multilingual Instruction Following
This work addresses the problem of evaluating multilingual instruction-following for LLM developers and researchers, providing a detailed benchmark but is incremental as it builds on existing evaluation frameworks.
The paper tackles the lack of systematic evaluation of large language models in multilingual instruction-following by introducing XIFBench, a benchmark with 558 instructions across six languages and five constraint categories, revealing performance disparities and insights into factors like language resources and cultural specificity.
Large Language Models (LLMs) have demonstrated remarkable instruction-following capabilities across various applications. However, their performance in multilingual settings lacks systematic investigation, with existing evaluations lacking fine-grained constraint analysis across diverse linguistic contexts. We introduce XIFBench, a comprehensive constraint-based benchmark for evaluating multilingual instruction-following abilities of LLMs, comprising 558 instructions with 0-5 additional constraints across five categories (Content, Style, Situation, Format, and Numerical) in six languages spanning different resource levels. To support reliable and consistent cross-lingual evaluation, we implement three methodological innovations: cultural accessibility annotation, constraint-level translation validation, and requirement-based evaluation using English requirements as semantic anchors across languages. Extensive experiments with various LLMs not only quantify performance disparities across resource levels but also provide detailed insights into how language resources, constraint categories, instruction complexity, and cultural specificity influence multilingual instruction-following. Our code and data are available at https://github.com/zhenyuli801/XIFBench.