Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- en
|
| 4 |
+
pipeline_tag: text-to-video
|
| 5 |
+
tags:
|
| 6 |
+
- video-generation
|
| 7 |
+
- world-model
|
| 8 |
+
- pytorch
|
| 9 |
+
- dit
|
| 10 |
+
library_name: pytorch
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
# HyDRA: Out of Sight but Not Out of Mind: Hybrid Memory for Dynamic Video World Models
|
| 14 |
+
|
| 15 |
+
This is the official Hugging Face model repository for **HyDRA** (Hybrid Memory for Dynamic Video World Models).
|
| 16 |
+
|
| 17 |
+
π **GitHub Repository:** [H-EmbodVis/HyDRA](https://github.com/H-EmbodVis/HyDRA)
|
| 18 |
+
π **Project Page:** [Hybrid-Memory-in-Video-World-Models](https://kj-chen666.github.io/Hybrid-Memory-in-Video-World-Models/)
|
| 19 |
+
|
| 20 |
+
## π Overview
|
| 21 |
+
|
| 22 |
+
While recent video world models excel at simulating static environments, they share a critical blind spot: the physical world is dynamic. When moving subjects exit the camera's field of view and later re-emerge, current models often lose track of them.
|
| 23 |
+
|
| 24 |
+
To bridge this gap, we introduce **Hybrid Memory**, a novel paradigm that requires models to simultaneously act as precise archivists for static backgrounds and vigilant trackers for dynamic subjects. **HyDRA** is a specialized memory architecture that compresses contexts into memory tokens and utilizes a spatiotemporal relevance-driven retrieval mechanism.
|
| 25 |
+
|
| 26 |
+
## π― Task & Capabilities
|
| 27 |
+
- **Task:** Text-to-Video Generation / Video World Modeling
|
| 28 |
+
- **Input:** Text prompts, camera poses, and initial video latents.
|
| 29 |
+
- **Output:** High-fidelity video sequences maintaining both identity and motion continuity of dynamic subjects, even during out-of-view intervals.
|
| 30 |
+
|
| 31 |
+
## π Usage
|
| 32 |
+
|
| 33 |
+
To use these weights, please refer to our GitHub repository: [H-EmbodVis/HyDRA](https://github.com/H-EmbodVis/HyDRA)
|
| 34 |
+
|
| 35 |
+
|
| 36 |
+
## π Citation
|
| 37 |
+
If you find our work useful, please consider citing:
|
| 38 |
+
|
| 39 |
+
```bibtex
|
| 40 |
+
@article{chen2026out,
|
| 41 |
+
title = {Out of Sight but Not Out of Mind: Hybrid Memory for Dynamic Video World Models},
|
| 42 |
+
author = {Chen, Kaijin and Liang, Dingkang and Zhou, Xin and Ding, Yikang and Liu, Xiaoqiang and Wan, Pengfei and Bai, Xiang},
|
| 43 |
+
journal = {arXiv preprint arXiv:2603.25716},
|
| 44 |
+
year = {2026}
|
| 45 |
+
}
|