Reproduce “AutoAdapt: Reliable Few-Shot Adaptation under Clinical Distribution Shifts” 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
Song Wang, Jie Peng, Davis Hobley, Zachary Plotkin, Tianlong Chen
AutoAdapt treats few-shot clinical adaptation as choosing complete model recipes and deciding whether patient evidence supports that choice. It compares recipes using patient-held-out predictions and reliability signals, combines supported options, and retains a default when evidence is weak. On MIMIC-III→ALOTT, its six-task mean patient-level AUROC is .701 versus .656 for LoraHub, leading on five tasks; OTTEHR leads on oxygenation crisis.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.