Update task category and add license

This PR updates the `task_categories` to more accurately reflect the primary focus of the DORI benchmark, which is evaluating object orientation understanding in MLLMs. It also adds the license information. Please confirm the license is MIT.

README.md

@@ -1,10 +1,9 @@
 ---
-task_categories:
-- visual-question-answering
-- robotics
-- multiple-choice
 language:
 - en
+task_categories:
+- image-text-to-text
+pretty_name: DORI
 tags:
 - vlm
 - multimodal
@@ -16,7 +15,7 @@ tags:
 - benchmark
 - reasoning
 - vision
-pretty_name: DORI
+license: mit
 ---
 
 ## Dataset Details
@@ -31,7 +30,7 @@ What makes DORI unique is its systematic approach to isolating orientation perce
 
 - **Curated by:** Anonymous Authors 
 - **Language(s) (NLP):** English
-- **License:** Not specified in the current documentation
+- **License:** MIT License
 
 ### Dataset Sources
 
@@ -107,7 +106,7 @@ The distribution of tasks in the dataset is:
 - Viewer-Scene Direction: 20%
 - Inter-Object Direction: 19%
 - View Parallelism: 10%
-- Single-Axis Rotation: 9%
+- Single-axis Rotation: 9%
 - Directional Facing: 9%
 - Canonical Orientation: 5%
 
@@ -228,4 +227,4 @@ Anonymous Authors
 
 ## Dataset Card Contact
 
-For more information, please contact the authors 
+For more information, please contact the authors