MedHal addresses critical gaps in medical AI safety by providing a comprehensive dataset specifically designed to detect hallucinations in medical texts.

Key Features:

• Large-scale dataset overcoming limitations of existing small medical datasets
• Domain-specific approach tailored for medical content evaluation
• Multi-task capability beyond single tasks like Question Answering
• Security-focused design to reduce harmful AI outputs in healthcare

This research is vital for healthcare applications where AI hallucinations can have serious consequences, potentially leading to patient harm. MedHal enables more reliable testing and improvement of AI systems for clinical decision support.

MedHal: An Evaluation Dataset for Medical Hallucination Detection