CIELAB is only approximately perceptually uniform, and the historical fix was non-Euclidean colour-difference formulas culminating in CIEDE2000

The ProLab paper (arXiv:2012.07653) states the uniformity target and CIELAB's shortfall in one primary, Tier-1 passage: a colour coordinate space is "perceptually uniform" when Euclidean distances in it correspond to perceived colour differences, but "the CIELAB space is only approximately uniform ... non-Euclidean colour difference formulas were being developed ... The successful outcome of these efforts was the CIEDE2000 formula." The significant structural point is where the correction was applied: rather than redesigning the coordinate space, colorimetry kept CIELAB's coordinates and layered a non-Euclidean distance function on top of them. CIEDE2000 is that distance function — a metric patch, not a new space.

This supplies a primary-source anchor for the more general fact that claim-cielab-designed-to-remedy-ciexyz-perceptual-nonuniformity set a uniformity goal CIELAB does not fully meet. It is adjacent to, but narrower in scope than, claim-cielab-lacks-perceptual-uniformity-in-blue-hues: ProLab documents the general residual non-uniformity and the non-Euclidean fix, but does not single out the blue region, so the blue-hue-specific claim keeps its own [unverified-claim] flag and its verification thread question-verify-cielab-perceptual-uniformity-and-blue-hue-nonuniformity.

The "keep the coordinates, reweight the metric" move is the reason CIELAB bridges the vault's gradient-geometry cluster rather than its sampling-artifact cluster — see observation-cielab-nonuniformity-bridges-gradient-geometry-not-sampling-artifact.