Skip to main content
. 2026 Jul 29;8:1900115. doi: 10.3389/fdgth.2026.1900115

Table 13.

Post hoc pseudo-client comparison of fixed inference, local reconstruction adaptation, and federated reconstruction adaptation.

Setting Accuracy Macro-F1 SVEB F1 VEB F1 Recon. MSE
No adaptation 93.94% 76.18% 51.60% 80.01% –
Local reconstruction adaptation 94.14% 76.45% 51.51% 80.78% 0.6426→0.6001
Federated reconstruction adaptation 93.98% 76.22% 51.42% 80.29% 0.6376→0.6224