Adjoint Inversion Reveals Holographic Superposition and Destructive Interference in CNN Classifiers

ArXi:2604.27529v1 Announce Type: new A foundational assumption in CNN interpretability -- that deep encoders suppress background pixels while classifiers merely select from a cleaned feature pool (the Spatial Funnel Hypothesis) -- remains untested due to spatial hallucinations in existing visualization tools. We address this by Using this framework as a geometric probe, we uncover the first pixel-level evidence of strong superposition in vision encoders.