CCD: Mitigating Hallucinations in Radiology MLLMs via Clinical Contrastive Decoding
CCD is a training-free and retrieval-free inference framework that integrates structured clinical signals from taskโspecific radiology expert models to mitigate hallucinations in radiology MLLMs.
Sep 27, 2025
CCD
Clinical Contrastive Decoding, a training-free and retrieval-free inference framework for radiology MLLMs.
Sep 27, 2025
๐จ Preprint out โ Clinical Contrastive Decoding๏ผ
A technique to mitigate hallucination in radiology MLLMs.
Sep 27, 2025
RadEval: A framework for radiology text evaluation
RadEval, an open-source framework for evaluating radiology texts, integrating diverse metrics and a richly annotated expert dataset to enhance reproducibility and benchmarking in radiology report generation.
Sep 22, 2025
๐ Paper Accepted at PSB 2026!
Our paper is accepted for poster presentation.
Sep 9, 2025
๐ Paper Accepted โ See You at EMNLP 2025!
Our paper has been accepted for System Demo with Oral.
Sep 9, 2025
RadEval
A framework for radiology text evaluation, focused on chest X-ray report generation.
Jul 4, 2025
๐ Paper Accepted โ See You at ACL 2025!
Our paper is accepted. We're heading to Vienna for ACL 2025!
May 15, 2025
Libra - What about next?๐ธ
Beyond temporal comparison - exploring future directions in radiology AI modeling
Apr 25, 2025
๐ค Talk & Poster at HealTAC 2025!
Our work is recommended for HealTAC 2025 in Glasgow.
Apr 22, 2025