Reproduce “Diagnostic Loss: Asymmetric Supervision for Radiology Report Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ignacy Kolton, Wenxuan Li, Zongwei Zhou, Jarosław B. Ćwikła, Arkadiusz Sitek, Pedro R. A. S. Bassi, Szymon Płotka
DLoss changes training for abdominal CT report generation to reward supported diagnoses and discourage unsupported ones. It pools probabilities across report positions for supported and unsupported concepts, without changing inference. Across six models, mean GREEN gains were 4.05 points on Merlin and 4.28 externally; complete RadGraph-F1 gains were 3.82 and 2.96 points. These results cover abdominal CT, not clinical safety.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.