RoAR-VAE: Improving Latent Diffusability through Robust Prediction
An effective visual tokenizer must preserve image fidelity while producing latents that downstream diffusion models can learn efficiently. Prior work improves latent diffusability through semantic alignment or joint tokenizer--diffusion training. Revisiting REPA-E, a representative approach, we find that removing its diffusion objective and dedicated generative blocks while retaining semantic prediction from corrupted latents yields comparable downstream generation performance under a matched setting. This observation motivates a general principle of robust prediction: generation-relevant information should remain recoverable from corrupted latents. We instantiate this principle with two ...
| Authors | Yang Yue, Fangyun Wei, Yitong Wang et al. |
|---|---|
| Primary category | cs.CV |
| Published | 19 October 2026 |
| Last updated at source | 19 October 2026 |
Popularity, last 2 days
0 now. Counts as reported by the source, snapshotted daily by Silytics.
Recent changes
- 20 Oct 2026 Description: An effective visual tokenizer must preserve image fidelity while producing latents that downstream diffusion models can learn efficiently. Prior work improves latent diffusability through semantic alignment or joint tokenizer--diffusion tra → An effective visual tokenizer must preserve image fidelity while producing latents that downstream diffusion models can learn efficiently. Prior work improves latent diffusability through semantic alignment or joint tokenizer--diffusion tra
- 20 Oct 2026 Topics added cs.CV removed hf-daily-papers, visual tokenizer, latent diffusability, semantic alignment, joint tokenizer-diffusion training, repa-e, diffusion objective, generative blocks, corrupted latents
Changes detected by Silytics between refreshes of the source.
Silytics shows public metadata from arXiv and links to the original. Information may be incomplete or out of date. Last refreshed 21 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.