TABLE 3.

Paired t-test $p$ -values comparing our model with top-performing baselines across datasets. Bold values indicate statistically significant improvements $(p<0.05)$ .

Comparison model	PMC-15M	NIH ChestX-ray14	IU X-ray	VinDr-CXR
WinCLIP	0.014	0.021	0.018	0.025
DRAEM	0.007	0.009	0.004	0.012
STPM	0.010	0.016	0.019	0.028
PaDiM	0.023	0.033	0.027	0.045

The values in bold are the best values.