Table 2.

Performance on the internal merged test set.

Model	Acc.	Macro-F1	PR-AUC	MCC
CNN ensemble baseline	99.74	99.74	99.95	0.996
ViT ensemble baseline	99.64	99.64	99.90	0.995
NeuroTrustNet	99.31	99.31	99.50	0.991
AdaptoVision baseline	88.04	88.15	93.10	0.843
RapidNet baseline	90.00	90.00	94.50	0.870

Accuracy, Macro-F1, and PR-AUC are reported in %, while MCC is reported in [−1, 1]. Bold values highlight the proposed NeuroTrustNet/AAS configuration.