Update task category and add relevant tags for MMInA dataset

This PR updates the `task_categories` metadata to `image-text-to-text` to better reflect the multimodal nature of the MMInA dataset, which involves processing both image and text inputs for web agent tasks.

Additionally, relevant tags (`web-agent`, `multihop`, `multimodal`, `web-navigation`, `benchmark`) have been added to improve the dataset's discoverability and categorization on the Hugging Face Hub.

README.md

@@ -1,16 +1,20 @@
 ---
-task_categories:
-- text-generation
 language:
 - en
 size_categories:
 - 1K<n<10K
+task_categories:
+- image-text-to-text
+tags:
+- web-agent
+- multihop
+- multimodal
+- web-navigation
+- benchmark
 ---
 
 # MMInA: Benchmarking Multihop Multimodal Internet Agents Dataset
 
-
-
 MMInA is a multihop and multimodal benchmark to evaluate the embodied agents for compositional Internet tasks.
 <img src="teaser.png" width=1000>
 
@@ -87,4 +91,4 @@ This dataset has the following limitations:
         archivePrefix={arXiv},
         primaryClass={cs.CV}
   }
-```
+```