Preprint
Aug 2026
Positive-Unlabeled Preference Optimization For Chest X-ray Report Generation
This work proposes PU-DPO, a preference optimization framework to prevent omission noise from corrupting the preference signal, and reformulates the objective under a positive-unlabeled (PU) learning framework, treating absent mentions as unlabeled rather than truly negative.
Y. Kobayashi, P. Ramesh, Muhammad Ahmed Chaudhry et al.
· 0 citations