Learning When to Trust Geometry: Probe-Relative Source Preference for Camouflaged Object Detection

Adaptive fusion assumes a gate can learn where one representation is more trustworthy than another. We test that assumption. We supervise a spatial gate with the detached pixelwise loss difference between two auxiliary segmentation probes—the exact minimizer of an entropy-regularized two-source risk—and instantiate it in \method, a camouflaged object detector fusing frozen SAM 2.1 semantics with frozen Depth Anything V2 geometry. is strong: after SHA-256 auditing removes seven contaminated training images, three seeds reach $S_{\alpha }$ of 0.9402, 0.9174, 0.9264, and 0.9327 on CHAMELEON, CAMO, COD10K, and NC4K, and matched-protocol inference beats ZoomNeXt-B4 on all four datasets with non-overlapping paired bootstrap intervals in $S_{\alpha }$, $F_{\beta }^w$, and $E_{\varphi }$. The gate nevertheless fails to identify trustworthy geometry: AUROC for the probe-preference event is 0.6057–0.6429, and as a selector it incurs 2.8–4.5$\times$ the regret of the trivial rule “always trust geometry.” We trace this to the target, not the gate. Because both probes are accurate, $|e_s-e_g|\le 0.05$ at 75.9–81.9% of pixels, which at $\tau =0.25$ confines the target to $[0.450,0.550]$ and leaves the gate a near-constant signal. Three instruments confirm the mechanism: the margin distribution; a five-condition corruption ladder in which mean gate response is perfectly rank-correlated with induced damage ($\rho =-1.00$ on every dataset) and therefore blind wherever damage is small; and oracles showing 25.4–33.7% of output-space complementarity unclaimed. Margin-based competence supervision degenerates precisely as its experts co-improve—a general obstacle for loss-difference routing, not an artifact of one architecture.