Pixel Space Doesn't Save You: Two Mechanisms of Conditioning Homogenization in Diffusion Transformers

A key motivation for pixel-space Diffusion Transformers is that deleting the VAE removes a lossy representation bottleneck. We show this fails for the conditioning pathway: the bottleneck was never in the VAE. Across 12 checkpoints and 9 architecture families, pixel- and latent-space models interleave along the homogenization axis, with conditioning vectors degenerating toward near-identical representations across instances. We trace this to two mechanisms—a frequency-schedule mismatch in pooled conditioning, and text-stream writeability in token conditioning—and derive from the first a training-free, checkpoint-level diagnostic, the norm ratio $\Vert t_{\mathrm{emb}}\Vert /\Vert y_{\mathrm{emb}}\Vert$ ($r=0.83$, $p<10^{-3}$), validated out-of-sample on four held-out architectures under pre-registered predictions. The same ratio predicts class-code fragility across those checkpoints ($r=-0.888$), showing functional consequences in models we did not train. We validate the second mechanism by perturbing only the prompt-specific residual, which changes the named attribute in 92–100% of trials on SD3.5 and FLUX. We then propose Conditioning Norm Balancing (CNB), a training-time regularizer that restores balanced conditioning geometry, and evaluate it under a shared-noise paired protocol (10,000 pairs per cell, McNemar's exact test, FID and accuracy from the same samples). CNB holds the ratio at parity and improves both metrics at each model's own optimal guidance, on both sides of the pixel/latent divide: $+2.70pp$ top-1 on SiT-B/2 ($p=1.8\times 10^{-10}$) and $+4.80pp$ on PixelDiT-B, which has no VAE ($p=2.6\times 10^{-20}$), with FID improving in both cases. At lower guidance, the two metrics give opposite verdicts on identical weights. A single FID number cannot certify a conditioning pathway.