Image and text do not share a coordinate system

In Kimi K3, a caption's embedding points almost perfectly backwards from its own image That's not a bug. It's a coordinate system choice, and you can undo it with a single rotation🧵

• a linear classifier separates image from text embeddings with 100% accuracy in all 3 models we tested (Kimi K3, Inkling, Qwen3-Omni). It just reads off each model's coordinate conventions, not what the modalities actually share. (full writeup with interactive figures is atWhere the conventions come from: each modality is packed in its own cone. Kimi K3 packs all text into 1.6°. Qwen3-Omni does the same to images (2.0°). Same architecture family, opposite choice. The encoder-free Inkling from @thinkymachineskeeps both wide.So we tried the dumbest possible fix: whiten each modality, then fit ONE rank-32 rotation on training pairs. No scaling. No translation. No MLP. A rotation is not allowed to move or reshape anything, so it can only turn the image basis into the text basis.Result: held-out images retrieve their own captions at 9.9x chance in Kimi K3 9.2x in Qwen3-Omni 3.2x in Inkling A pure rotation, fitted on 96 pairs, recovers that much of the "modality gap".If you compare image and text activations directly using probes, similarity metrics, steering, the native coordinates will lie to you. Part of the "gap" is basis mismatch, not missing content. --- This research is done at@whitecircle-- we are hiring!