Answers are grounded in this paper and cite their sources.
Accuracy reported in one sensor basis can hide a diagnosis that changes when the same record is re-expressed. We introduce an output-level O ( 3 ) audit that separates coordinate consistency, cross-band information retention, and physical-unit utility. MUET provides the primary audit; MCC5 checks operating profiles at two sensor locations on one rig, using the same 240 files. Across the fixed 1072-view tensor orbit, ordered coordinates had mean worst-case probability drift 0.5180, against 7.68 × 10 − 16 (maximum 1.94 × 10 − 15 ) for the proposed signature. Both collision witnesses passed. With six matched-severity blocks, MUET utility remained inconclusive (balanced accuracy 0.4167, 95% cluster interval 0.1667–0.6667). MCC5 exposed profile-specific utility shortfalls. The audit exposes a failure mode that accuracy cannot see and tells deployment teams whether to correct coordinates or gather stronger diagnostic evidence.
No takes yet. Share an insight, caveat, or question.
Zongyuan Yang (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: