Improve model card: Add pipeline tag, license, and enhance overview

README.md

@@ -1,13 +1,36 @@
 ---
 library_name: transformers
-tags: []
+tags:
+- gui-agent
+- vlm
+- reinforcement-learning
+- gui-grounding
+- gui-automation
+pipeline_tag: image-text-to-text
+license: apache-2.0
 ---
 
-# Introduction
+# GTA1: GUI Test-time Scaling Agent
+
+This repository contains the model presented in [GTA1: GUI Test-time Scaling Agent](https://huggingface.co/papers/2507.05791).
+
+<div align="center">
+  <img style="width: 100%" src="https://raw.githubusercontent.com/Yan98/GTA1/main/assets/img/model.png">
+</div>
+
+Graphical user interface (GUI) agents autonomously operate across platforms (e.g., Linux) to complete tasks by interacting with visual elements. This paper introduces **GTA1**, a GUI Test-time Scaling Agent, which addresses two main challenges: i) resolving ambiguity in task planning by introducing a test-time scaling method that samples and selects optimal action proposals; and ii) accurately grounding actions in complex, high-resolution interfaces. GTA1 proposes a model for improved grounding accuracy through reinforcement learning, leveraging inherent objective alignments that reward successful interactions.
+
+This model specifically focuses on the GUI grounding component of GTA1.
+
+**[Code on GitHub](https://github.com/Yan98/GTA1)** | **[Paper](https://huggingface.co/papers/2507.05791)** | **[Blog](https://huggingface.co/blog/HelloKKMe/grounding-r1)**
+
+---
+
+## About the Grounding Model
 
 Reinforcement learning (RL) (e.g., GRPO) helps with grounding because of its inherent objective alignment—rewarding successful clicks—rather than encouraging long textual Chain-of-Thought (CoT) reasoning. Unlike approaches that rely heavily on verbose CoT reasoning, GRPO directly incentivizes actionable and grounded responses. Based on findings from our [blog](https://huggingface.co/blog/HelloKKMe/grounding-r1), we share state-of-the-art GUI grounding models trained using GRPO.
 
-# Performance
+## Performance
 
 We follow the standard evaluation protocol and benchmark our model on three challenging datasets. Our method consistently achieves the best results among all open-source model families. Below are the comparative results:
 
@@ -36,7 +59,7 @@ We follow the standard evaluation protocol and benchmark our model on three chal
 > - UI-TARS-1.5 7B, Qwen2.5-VL-32B-Instruct, and Qwen2.5-VL-72B-Instruct are applied as our baseline models.
 > - ∆ indicates the performance improvement (∆) of our model compared to its baseline.
 
-# Inference
+## Inference
 Below is a code snippet demonstrating how to run inference using a trained model.
 
 ```python
@@ -125,4 +148,28 @@ pred_y*=scale_y
 print(pred_x,pred_y)
 ```
 
-Refer to our [code](https://github.com/Yan98/GTA1) for more details.
+## Agent Performance
+
+Refer to our [code](https://github.com/Yan98/GTA1) for more details on agent inference.
+
+## Contact
+
+Please contact `yan.yang@anu.edu.au` for any queries.
+
+## Acknowledgement
+
+We thank the open-source projects: [VLM-R1](https://github.com/om-ai-lab/VLM-R1), [Jedi](https://github.com/xlang-ai/OSWorld/blob/main/mm_agents/jedi_7b_agent.py), and [Agent-S2](https://github.com/simular-ai/Agent-S).
+
+## Citation
+If you use this repository or find it helpful in your research, please cite it as follows:
+```bibtex
+@misc{yang2025gta1guitesttimescaling,
+      title={GTA1: GUI Test-time Scaling Agent}, 
+      author={Yan Yang and Dongxu Li and Yutong Dai and Yuhao Yang and Ziyang Luo and Zirui Zhao and Zhiyuan Hu and Junzhe Huang and Amrita Saha and Zeyuan Chen and Ran Xu and Liyuan Pan and Caiming Xiong and Junnan Li},
+      year={2025},
+      eprint={2507.05791},
+      archivePrefix={arXiv},
+      primaryClass={cs.AI},
+      url={https://arxiv.org/abs/2507.05791}, 
+}
+```