The BiasMD dataset is a comprehensive resource designed to evaluate and mitigate biases in health-related responses generated by Large Language Models (LLMs). It contains 6,007 question-answer pairs that address ethically sensitive healthcare scenarios across diverse demographic factors. This dataset is part of a broader effort to enhance the ethical and accurate use of AI in healthcare communication.
