Uncertainty-Aware Deepfake Detection via Multi-View Structural Learning
An uncertainty-aware deepfake detection framework that identifies manipulations through inconsistencies across complementary evidence sources by introducing Inter-Branch Disagreement Calibration (IBDC), a disagreement-aware uncertainty modeling mechanism that links predictive uncertainty to conflicts among evidence streams.