Mila and the Queensland University of Technology (QUT) present AIMS.au, a specialized dataset to support the fine-tuning and evaluation of large language models for analyzing corporate modern slavery statements.

This dataset features annotated modern slavery statements from the Australian Modern Slavery Register. Designed to advance research in natural language processing applications in corporate compliance with business and human rights legislation, the collection enables more accurate assessment of how companies report on modern slavery risks and their compliance with the Modern Slavery Act. This work has been published at the International Conference on Learning Representations (ICLR) 2025. All associated files and code are openly available on Hugging Face, GitHub, and Figshare. You can access the paper on arXiv.