Add paper link, project page, and GitHub repository

This PR improves the dataset card by:
1. Adding a link to the Cosmos 3 technical report.
2. Adding links to the official project page and GitHub repository.
3. Adding the `any-to-any` task category to the metadata to reflect the omnimodal nature of the research this dataset supports.

README.md

@@ -2,24 +2,30 @@
 language:
 - en
 license: other
+pretty_name: PhysicalAI-SDG-SynHuman
 license_name: nvidia-sdg-synhuman-license
 license_link: https://huggingface.co/datasets/nvidia/PhysicalAI-SDG-SynHuman/resolve/main/LICENSE
-pretty_name: PhysicalAI-SDG-SynHuman
+task_categories:
+- any-to-any
 tags:
 - video
 - synthetic
 - human
 - physical-ai
+- world-model
 ---
 
-
 <p align="center">
   <img src="assets/hugging_face_gif_01_small.gif" alt="SDG-SynHuman Preview" width="100%">
 </p>
 
+# SDG-SynHuman
+
+[**Project Page**](https://research.nvidia.com/labs/cosmos-lab/cosmos3) | [**Paper**](https://huggingface.co/papers/2606.02800) | [**GitHub**](https://github.com/nvidia/cosmos)
+
 ## Dataset Description: <br>
 
-The SDG-SynHuman is a large-scale synthetic video dataset of digital humans rendered in diverse indoor and outdoor 3D environments. The dataset contains 236,937 clips, totaling approximately 5,841 hours of video, and is designed to support training and post-training of NVIDIA Cosmos world foundation models and related physical AI research.
+The SDG-SynHuman is a large-scale synthetic video dataset of digital humans rendered in diverse indoor and outdoor 3D environments. The dataset contains 236,937 clips, totaling approximately 5,841 hours of video, and is designed to support training and post-training of NVIDIA Cosmos world foundation models and related physical AI research as presented in [Cosmos 3: Omnimodal World Models for Physical AI](https://huggingface.co/papers/2606.02800).
 
 Each sample is a temporally coherent 60-120 second video clip rendered at 1080p and 30 fps. Clips contain multiple digital humans performing animation sequences in a sampled 3D environment with a controlled camera trajectory. The dataset spans 4,050 digital human assets, 8,184 unique animations, 198 indoor environments, 200 outdoor city environments, and 14 camera-motion scenarios, providing broad variation in human appearance, motion, scene context, lighting, and camera behavior.