SOURCE-LINKED INTELLIGENCE

Calibration Count Reuse: Validity Does Not Determine Efficiency

arXiv · AI, language, vision and robotics · article · Sep 21, 2026 · UTC

Calibration count reuse raises separate validity and efficiency questions. We give a validity criterion for general count-dependent nonconformity scores: transferring one count from another class to the scored class must not improve its conformity. A leave-self-out full conformal reference proves the criterion without requiring normalization or preservation of same-class score order. For a common separable transformation, universal exchangeable validity is equivalent to being nondecreasing in the count, provided $Kα\geq 1$; normalized multiplicative weights obey the complementary nonincreasing

Read original source ↗ Open in workspace

recordType

paper

region

Global