Available files:

iu-split-<split>.txt: list of reports per split (train, val, test)
<dataset>-broken.txt: list of broken images from each dataset
mimic-expert*.txt: expert evaluation of ~1000 samples, see our AIM paper