Add task categories and paper link

This PR adds the missing `task_categories` metadata (`image-text-to-text` and `video-text-to-text`) so the dataset is correctly categorized, and links the dataset card to its paper on the Hugging Face Hub.

README.md

@@ -2,6 +2,9 @@
 license: other
 license_name: nvidia-evaluation-data-license
 license_link: LICENSE.md
+task_categories:
+- image-text-to-text
+- video-text-to-text
 dataset_info:
   splits:
   - name: test
@@ -12,6 +15,8 @@ dataset_info:
 
 *Video ANalysis Tasks Across Generalized Environments*
 
+**Paper:** [VANTAGE-Bench: Evaluating the Infrastructure AI Gap in Vision-Language Models](https://huggingface.co/papers/2609.09396)
+
 ## Dataset Description
 
 VANTAGE-BENCH is the first public benchmark purpose-built for evaluating visual understanding on video captured by fixed infrastructure cameras. It spans three real-world domains — warehouse, smart city / Intelligent Transportation Systems (ITS), and smart spaces — across six spatio-temporal video understanding tasks including video question answering (VQA), temporal grounding, dense video captioning, event verification, spatial grounding, and spatio-temporal tracking.
@@ -148,4 +153,4 @@ Please report quality, risk, security vulnerabilities or NVIDIA AI Concerns [her
 
 ## Changelog
 
-See [CHANGELOG.md](./CHANGELOG.md) for release history.
+See [CHANGELOG.md](./CHANGELOG.md) for release history.