Add video-text-to-text task category, MIT license, and link VG-GUI-Bench paper

README.md

@@ -74,6 +74,9 @@ configs:
     path: data/test-*
   - split: test_unseen_os
     path: data/test_unseen_os-*
+task_categories:
+- video-text-to-text
+license: mit
 ---
 
 [Paper](https://arxiv.org/abs/2505.12632) |
@@ -88,6 +91,8 @@ _MONDAY_ (Mobile OS Navigation Task Dataset for Agents from YouTube) is a cross-
 - **333K** detected scenes, each representing a temporally segmented step within a mobile navigation task.
 - **313K** identified actions, including touch, scroll, hardware, typing, long press, multi touch and zoom.
 
+MONDAY is also used to construct **VG-GUI-Bench** in the ECCV 2026 paper [Bridging VideoQA and Video-Guided Agentic Tasks via Generalized Keyframe Extraction](https://huggingface.co/papers/2606.29445) ([Code](https://github.com/VG-GUI-TASKER/VG-GUI-TASKER)).
+
 Please visit our [project page](https://monday-dataset.github.io/) for more details.
 
 ## Data Fields
@@ -159,4 +164,11 @@ Please visit our [project page](https://monday-dataset.github.io/) for more deta
     booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
     year={2025}
 }
-````
+
+@inproceedings{fan2026bridging,
+  title     = {Bridging VideoQA and Video-Guided Agentic Tasks via Generalized Keyframe Extraction},
+  author    = {Fan, Sunqi and Liu, Qingle and Yin, Runqi and Guo, Meng-Hao and Yang, Shuojin},
+  booktitle = {European Conference on Computer Vision (ECCV)},
+  year      = {2026}
+}
+```