The Shift–Utility Gap: Representation Shift Does Not Determine Predictive Utility in 3D Molecular Prediction

Three-dimensional (3D) molecular models can exploit geometric information beyond coordinate-independent two-dimensional (2D) topology, but they also inherit a deployment vulnerability: predictions depend on the coordinate realization supplied at inference time. We study a problem distinct from representation-shift detection itself. A strongly shifted 3D representation can still outperform a 2D expert for a particular molecule, while an apparently in-distribution representation need not provide additional predictive value. We call this mismatch the shift–utility gap. On OpenADMET-ExpansionRx, end-to-end Uni-Mol falls from macro $R^2=0.5292$ on original coordinates to $-0.4738$ after coordinate randomization; a matched Frad analysis drops from $0.360646$ to $-3.422540$. Across six external regression datasets, clean-versus-random shift is detected almost perfectly (AUROC 0.9957–1.0000), whereas clean sample-level utility discrimination remains near chance (AUROC 0.4430–0.5256) and randomized-condition utility varies substantially (0.4206–0.6539). We introduce CORE (COordinate-shift-aware Routing of Experts), which combines shift evidence with 2D/3D predictions, disagreement, endpoint context, and a bounded 2D-anchored robust path. Under randomized coordinates, CORE reaches macro $R^2=0.5710$ versus $0.5465$ for endpoint-fixed static mixing (BH-adjusted $q=0.0133$); under original coordinates static mixing is slightly higher ($0.6022$ versus $0.5943$), but not significantly so ($q=0.3256$). Full-architecture transfer on Lipo and ESOL shows the same qualitative failure-containment pattern beyond OpenADMET. The main contribution is therefore the empirical distinction between representation shift and relative predictive utility; CORE is one selective-routing response to that distinction.