VLM3D Tasks

A four‑task benchmark on the CT‑RATE dataset pushing 3‑D vision‑language AI. Tasks cover (1) radiology report generation, (2) 18‑label abnormality classification, (3) self‑supervised localization of five key pathologies, and (4) text‑guided synthesis of realistic chest CT volumes.