Add paper link, GitHub repository, and task categories

Hi! I'm Niels from the Hugging Face community science team. I've updated the dataset card to include:
- A link to the paper and GitHub repository.
- The `image-text-to-text` task category in the metadata.
- A citation section to ensure the authors are properly credited.

These additions will make the dataset more discoverable and provide users with the necessary context for its use in multimodal social simulation research.

README.md

@@ -1,10 +1,12 @@
 ---
-license: mit
-pretty_name: AgentViSS
 language:
 - en
+license: mit
 size_categories:
 - n<1K
+task_categories:
+- image-text-to-text
+pretty_name: AgentViSS
 tags:
 - multimodal
 - social-simulation
@@ -23,6 +25,8 @@ This repository contains the public data package for **AgentViSS: Can Agents
 Read the Room? Benchmarking Visual Social Intelligence in Multimodal
 Simulation**.
 
+[**Paper**](https://huggingface.co/papers/2606.15152) | [**GitHub**](https://github.com/JunsWan/AgentViSS)
+
 ## Files
 
 ```text
@@ -118,3 +122,13 @@ You can also clone the repository:
 ```bash
 git clone https://huggingface.co/datasets/JunsWan/AgentViSS
 ```
+
+## Citation
+
+```bibtex
+@misc{agentviss2026,
+  title = {Can Agents Read the Room? Benchmarking Visual Social Intelligence in Multimodal Simulation},
+  author = {Anonymous Authors},
+  year = {2026}
+}
+```