DeltaTok
Collection
Pre-trained DeltaTok tokenizer, DeltaWorld predictor, and evaluation heads for segmentation, depth, and RGB reconstruction. See deltatok.github.io. • 6 items • Updated • 1
DeltaWorld (Predictor) — Kinetics-700
Generative ViT-B predictor that samples future delta tokens for diverse video prediction. Trained on Kinetics-700 at 512x512 resolution. Requires a trained DeltaTok tokenizer and a frozen DINOv3 ViT-B backbone (neither included).
Usage
See the DeltaTok GitHub repository for training and evaluation code.
Acknowledgements
Citation
@inproceedings{kerssies2026deltatok,
title = {A Frame is Worth One Token: Efficient Generative World Modeling with Delta Tokens},
author = {Kerssies, Tommie and Berton, Gabriele and He, Ju and Yu, Qihang and Ma, Wufei and de Geus, Daan and Dubbelman, Gijs and Chen, Liang-Chieh},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
year = {2026}
}
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support