Reproduction: Global Geometry Is Not Enough for Vision Representations