Anchoring LLM-based Chest X-ray Report Generation (NeurIPS 2026) by Liu ChangAnchoring LLM-based Chest X-ray Report Generation (NeurIPS 2026) by Liu Chang

Anchoring LLM-based Chest X-ray Report Generation (NeurIPS 2026)

Portfolio project
Liu Chang
Co-authored a NeurIPS 2026 poster on diffusion large language models for radiology report generation, namely, Anchor-Gen. This study uses a diffusion language planner that first builds a structured clinical plan; and then utilizes a gated bridge feeds the plan to the LLM decoder. Results on IU X-Ray, MIMIC-CXR, CXR-PRO w.r.t. NLG, CE metrics outperform existing methods. In this manuscript, I contributed to the full lifecycle from designing the idea from scratch to the final paper acceptance with my co-authors. Links will be updated as soon as the metadata is available.

1 like · 1 view · Oct 7, 2026

Co-authored a NeurIPS 2026 poster on diffusion large language models for radiology report generation, namely, Anchor-Gen.

Tags