M2Lingual: Enhancing Multilingual, Multi-Turn Instruction Alignment in Large Language Models

2024-06-24

arXiv.org

AI Summary

Existing instruction fine-tuning (IFT) datasets exhibit strong bias toward high-resource languages—especially English—and lack comprehensive multilingual and multi-turn task coverage, thereby hindering large language models’ (LLMs’) ability to achieve instruction alignment in low-resource languages. To address this, we introduce M2Lingual: the first fully synthetic, multilingual, multi-turn, task-oriented IFT dataset, spanning 70 languages, 17+ NLP tasks, and 182K high-quality samples. Our method introduces a novel two-step Evolution taxonomy to guide the generation of complex multi-turn instructions, integrating seed-guided evolution, multilingual rewriting, dialogue-structure modeling, and quality-controlled diversity sampling. Experiments demonstrate that M2Lingual substantially improves performance across LLMs of varying scales on multilingual and multi-turn understanding and generation tasks. The dataset and generation code are publicly released on Hugging Face and GitHub.

Technology Category

• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Instruction finetuning (IFT) is critical for aligning Large Language Models (LLMs) to follow instructions. While many effective IFT datasets have been introduced recently, they predominantly focus on high-resource languages like English. To better align LLMs across a broad spectrum of languages and tasks, we propose a fully synthetic, novel taxonomy (Evol) guided Multilingual, Multi-turn instruction finetuning dataset, called M2Lingual. It is constructed by first selecting a diverse set of seed examples and then utilizing the proposed Evol taxonomy to convert these seeds into complex and challenging multi-turn instructions. We demonstrate the effectiveness of M2Lingual by training LLMs of varying sizes and showcasing the enhanced performance across a diverse set of languages. We contribute the 2 step Evol taxonomy with the guided generation code: https://github.com/ServiceNow/M2Lingual, as well as the first fully synthetic, general and task-oriented, multi-turn, multilingual dataset built with Evol - M2Lingual: https://huggingface.co/datasets/ServiceNow-AI/ M2Lingual - containing 182K total IFT pairs, covering 70 languages and 17+ NLP tasks.

Problem

• Enhance multilingual instruction alignment in LLMs
• Address lack of IFT datasets for low-resource languages
• Improve multi-turn instruction handling across diverse tasks

Innovation

• Synthetic multilingual multi-turn instruction dataset
• Evol taxonomy for complex instruction generation
• Enhanced LLM performance across 70 languages