Investigating Multilingual Instruction-Tuning: Do Polyglot Models Demand for Multilingual Instructions?

The adaption of multilingual pre-trained LLMs into eloquent and helpful assistants is essential to facilitate their use across different language regions. In that spirit, we are the first to conduct an extensive study of the performance of multilingual models instruction-tuned on different language compositions on textit{parallel} textit{instruction-tuning} benchmarks across a selection of the most spoken Indo-European languages. We systematically examine the effects of language and instruction dataset size on a mid-sized and a large, multilingual LLMs by instruction-tuning them on parallel instruction-tuning datasets. Our results demonstrate that instruction-tuning on parallel instead of monolingual corpora benefits cross-lingual instruction following capabilities by up to 9.9%. Furthermore, we show that the Superficial Alignment Hypothesis does not hold in general, as the investigated multilingual 7B parameter model presents a counter-example requiring large-scale instruction-tuning datasets. Finally, we conduct a human annotation study to understand the alignment between human-based and GPT-4-based evaluation within multilingual chat scenarios.

  • Published in:
    Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing
  • Type:
    Inproceedings
  • Authors:
    Weber, Alexander Arno; Thellmann, Klaudia; Ebert, Jan; Flores-Herr, Nicolas; Lehmann, Jens; Fromm, Michael; Ali, Mehdi
  • Year:
    2024
  • Source:
    https://aclanthology.org/2024.emnlp-main.1159/

Citation information

Weber, Alexander Arno; Thellmann, Klaudia; Ebert, Jan; Flores-Herr, Nicolas; Lehmann, Jens; Fromm, Michael; Ali, Mehdi: Investigating Multilingual Instruction-Tuning: Do Polyglot Models Demand for Multilingual Instructions?, Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, 2024, https://aclanthology.org/2024.emnlp-main.1159/, Weber.etal.2024a,

Associated Lamarr Researchers

Prof. Dr. Jens Lehmann

Principal Investigator to the profile