This PR updates the dataset card to include:
- A link to the paper ([MVTamperBench: Evaluating Robustness of Vision-Language Models](https://arxiv.org/abs/2412.19794)).
- A link to the code repository.
- The correct `task_categories` entry (`video-classification`).
The `visual-question-answering` task category was removed because the abstract and description of the dataset clearly indicate a focus on video classification and manipulation detection within the context of Vision-Language Models.
