The Meta-Classifier acts as the final "Judge". It does not look at the video pixels; instead, it analyzes the numerical scores generated by all the independent physical and biological sensors.
Inputs: 12-14 distinct anomaly scores (0.0 to 1.0).
Self-Attention: Learns which sensors to trust based on the context (e.g. ignoring color anomalies if the video is black and white).
XAI Override: Hard-coded to automatically override the neural network and flag the video as a Deepfake if any critical biological sensor (like Geometry) exceeds 70% anomaly.