BioD2C introduces a novel dual-level semantic consistency framework that significantly improves how AI systems answer medical image-based questions.

• Creates stronger alignment between visual and textual medical information
• Introduces the new BioVGQ dataset for testing biomedical visual grounding capabilities
• Achieves state-of-the-art performance across multiple biomedical VQA benchmarks
• Enhances reasoning capabilities by enforcing consistency at both image-question and answer levels

This research directly impacts medical diagnostics assistive technology by improving AI systems' ability to understand and answer complex medical visual queries with greater accuracy and reliability.

BioD2C: A Dual-level Semantic Consistency Constraint Framework for Biomedical VQA