Reproduce “Ranking-Aware Prompt Optimization for Multimodal Clinical Diagnosis” 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
Tian Xia, Minghao Liu, Yiqing Liang, Laixi Shi, Jiayun Wang
Ranking-PE adapts reflective prompt search for chest X-ray diagnosis by replacing individual correctness events with positive–negative score comparisons, so candidate selection targets AUROC. Across three MIMIC tasks, it improves test AUROC over accuracy-based search. Qwen3-VL-8B with vision-encoder-tuned SFT averages 73.8% versus 68.0%. This retrospective study covers three diseases and two open-source model families, with small patient-disjoint test sets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.