Improve model card and add metadata

This PR improves the model card for the EAPO model. It adds:
- Relevant metadata: `pipeline_tag: image-text-to-text` and `library_name: transformers`.
- Links to the paper ([Learning to Explore: Scaling Agentic Reasoning via Exploration-Aware Policy Optimization](https://huggingface.co/papers/2605.08978)), the project page, and the official GitHub repository.
- A brief description of the model based on the abstract.
- The correct BibTeX citation.

README.md

@@ -1,3 +1,37 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+library_name: transformers
+pipeline_tag: image-text-to-text
+---
+
+# Learning to Explore: Scaling Agentic Reasoning via Exploration-Aware Policy Optimization (EAPO)
+
+[**📄 Paper**](https://arxiv.org/abs/2605.08978) | [**🌐 Website**](https://xingyuan-project.github.io/m2cl.github.io/) | [**💻 Code**](https://github.com/HansenHua/EAPO-ICML26)
+
+This repository contains the model weights for **EAPO**, presented in the paper [Learning to Explore: Scaling Agentic Reasoning via Exploration-Aware Policy Optimization](https://huggingface.co/papers/2605.08978), which was accepted at ICML 2026.
+
+## Introduction
+
+EAPO is an exploration-aware reinforcement learning framework that enables LLM agents to adaptively explore only when uncertainty is high. It introduces a fine-grained reward function via variational inference to evaluate exploratory actions and an exploration-aware grouping mechanism to separate exploratory actions from task-completion actions during optimization.
+
+The model is based on the **Qwen2.5-VL** architecture and demonstrates consistent improvements across challenging text-based and GUI-based agent benchmarks.
+
+## Resources
+
+- **Paper:** [arXiv:2605.08978](https://arxiv.org/abs/2605.08978)
+- **Repository:** [GitHub - HansenHua/EAPO-ICML26](https://github.com/HansenHua/EAPO-ICML26)
+- **Project Page:** [Website](https://xingyuan-project.github.io/m2cl.github.io/)
+
+## Citation
+
+If you find our paper and code useful in your research, please consider citing:
+
+```bibtex
+@inproceedings{
+hua2026learning,
+title={Learning to Explore: Scaling Agentic Reasoning via Exploration-Aware Policy Optimization},
+author={Xingyuan Hua and Sheng Yue and Ju Ren},
+booktitle={The Forty-third International Conference on Learning Representations},
+year={2026}
+}
+```