Improve dataset card for PAC Bench: Add paper, project page, and enhanced metadata

README.md

@@ -1,3 +1,23 @@
----
-license: mit
----
+---
+license: mit
+task_categories:
+- robotics
+tags:
+- benchmark
+- physical-reasoning
+- vlm
+- robot-manipulation
+---
+
+This repository contains PAC Bench, a comprehensive benchmark dataset designed to systematically evaluate Vision-Language Models (VLMs) on their understanding of core Properties, Affordances, and Constraints (PAC) from a task executability perspective in the context of robot manipulation.
+
+The dataset was presented in the paper: [PAC Bench: Do Foundation Models Understand Prerequisites for Executing Manipulation Policies?](https://huggingface.co/papers/2506.23725)
+
+**Project Page:** [https://pacbench.github.io/](https://pacbench.github.io/)
+
+PAC Bench features a diverse dataset with over 30,000 annotations, comprising:
+*   673 real-world images (115 object classes, 15 property types, and 1 to 3 affordances defined per class)
+*   100 real-world humanoid-view scenarios
+*   120 unique simulated constraint scenarios across four tasks.
+
+This benchmark aims to highlight limitations in current VLMs' grasp of fundamental physical concepts, guiding the development of more robust, physically grounded models for robotic applications.