Add paper, project page, and code links to dataset card

README.md

@@ -1,11 +1,34 @@
 ---
 license: cc-by-4.0
+task_categories:
+- image-to-video
 tags:
 - video
 - gameplay
 - rl
 - mario
 ---
-Expert gameplay video from Super Mario Bros, divided to episodes and train/val splits. Frames are 1280x720.
 
-Source: https://www.youtube.com/watch?v=c_b9Yn34pdI
+# Mario Gameplay Dataset
+
+Expert gameplay video from Super Mario Bros, divided into 100-frame episodes and train/val splits. Frames are 1280x720.
+
+This dataset was used to evaluate **Latent Particle World Models (LPWM)**, a self-supervised object-centric world model presented in:
+
+- **Paper:** [Latent Particle World Models: Self-supervised Object-centric Stochastic Dynamics Modeling](https://huggingface.co/papers/2603.04553)
+- **Project Page:** [https://taldatech.github.io/lpwm-web](https://taldatech.github.io/lpwm-web)
+- **GitHub Repository:** [https://github.com/taldatech/lpwm](https://github.com/taldatech/lpwm)
+- **Original Source:** [https://www.youtube.com/watch?v=c_b9Yn34pdI](https://www.youtube.com/watch?v=c_b9Yn34pdI)
+
+## Citation
+
+```bibtex
+@inproceedings{
+daniel2026latent,
+title={Latent Particle World Models: Self-supervised Object-centric Stochastic Dynamics Modeling},
+author={Tal Daniel and Carl Qi and Dan Haramati and Amir Zadeh and Chuan Li and Aviv Tamar and Deepak Pathak and David Held},
+booktitle={The Fourteenth International Conference on Learning Representations},
+year={2026},
+url={https://openreview.net/forum?id=lTaPtGiUUc}
+}
+```