ViTok-v2: Scaling Native Resolution Auto-Encoders to 5 Billion Parameters
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866910195210256384 |
|---|---|
| author | Hansen-Estruch, Philippe Chen, Jiahui Ramanujan, Vivek Zohar, Orr Ping, Yan Sinha, Animesh Georgopoulos, Markos Schoenfeld, Edgar Hou, Ji Juefei-Xu, Felix Vishwanath, Sriram Thabet, Ali |
| author_facet | Hansen-Estruch, Philippe Chen, Jiahui Ramanujan, Vivek Zohar, Orr Ping, Yan Sinha, Animesh Georgopoulos, Markos Schoenfeld, Edgar Hou, Ji Juefei-Xu, Felix Vishwanath, Sriram Thabet, Ali |
| contents | Vision Transformer (ViT) autoencoders have emerged as compelling tokenizers for images, offering improved reconstruction over convolutional tokenizers. However, existing ViT tokenizers cannot explore this landscape as performance degrades outside training resolutions, and reliance on adversarial losses prevents stable scaling. ViTok (Hansen-Estruch et al., 2025) found that the compression ratio r mediates a reconstruction-generation trade-off where lower r means better reconstructions but harder generations, so improving tokenizer reconstruction is key to more Pareto-optimal tokenizers. We introduce ViTok-v2, which addresses these limitations with native resolution support via NaFlex for generalization across resolutions and aspect ratios, and a novel DINOv3 perceptual loss that replaces both LPIPS and GAN objectives for stable training at any scale. ViTok-v2 is trained on about 2B images and scaled to 5B parameters, the largest image autoencoder to date. ViTok-v2 matches or exceeds state-of-the-art reconstruction at 256p and outperforms all baselines at 512p and above. In joint scaling experiments with flow matching generators, we show that scaling both the autoencoder and the generator advances the Pareto frontier of this trade-off. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_05331 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | ViTok-v2: Scaling Native Resolution Auto-Encoders to 5 Billion Parameters Hansen-Estruch, Philippe Chen, Jiahui Ramanujan, Vivek Zohar, Orr Ping, Yan Sinha, Animesh Georgopoulos, Markos Schoenfeld, Edgar Hou, Ji Juefei-Xu, Felix Vishwanath, Sriram Thabet, Ali Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning Vision Transformer (ViT) autoencoders have emerged as compelling tokenizers for images, offering improved reconstruction over convolutional tokenizers. However, existing ViT tokenizers cannot explore this landscape as performance degrades outside training resolutions, and reliance on adversarial losses prevents stable scaling. ViTok (Hansen-Estruch et al., 2025) found that the compression ratio r mediates a reconstruction-generation trade-off where lower r means better reconstructions but harder generations, so improving tokenizer reconstruction is key to more Pareto-optimal tokenizers. We introduce ViTok-v2, which addresses these limitations with native resolution support via NaFlex for generalization across resolutions and aspect ratios, and a novel DINOv3 perceptual loss that replaces both LPIPS and GAN objectives for stable training at any scale. ViTok-v2 is trained on about 2B images and scaled to 5B parameters, the largest image autoencoder to date. ViTok-v2 matches or exceeds state-of-the-art reconstruction at 256p and outperforms all baselines at 512p and above. In joint scaling experiments with flow matching generators, we show that scaling both the autoencoder and the generator advances the Pareto frontier of this trade-off. |
| title | ViTok-v2: Scaling Native Resolution Auto-Encoders to 5 Billion Parameters |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2605.05331 |