Reduces medical hallucinations and enhances clinical accuracy in radiology multimodal large language models.
Sponsored
Multimodal large language models (MLLMs) are advancing radiology by combining image and text understanding, but often generate inaccurate or unsupported clinical details—so-called medical hallucinations. Clinical Contrastive Decoding (CCD) is a training-free and retrieval-free inference framework that integrates structured clinical signals from task‑specific radiology expert models. CCD reduces these hallucinations and improves clinical accuracy without changing the base MLLM, offering a practical way to make radiology MLLMs more reliable.
Key Features
015 GitHub stars
02Compatible with Libra/LLaVA-architecture radiology MLLMs and multiple expert models
03Enhances clinical accuracy without modifying the base MLLM architecture
04Training-free and retrieval-free inference framework
05Integrates structured clinical signals from radiology expert models
06Mitigates medical hallucinations in MLLM-generated radiology reports
Use Cases
01Provides an educational tool for demonstrating medical image interpretations and temporal progressions
02Supports radiologists in drafting and reviewing chest X-ray reports for clinical decision making
03Serves as a research utility for investigating automated report generation and visual-language alignment in medical imaging
