Add model card and metadata

This PR adds a model card for the EnsemJudge detector (LoRA adapter for Qwen2.5-7B-Instruct). It populates the YAML metadata with the appropriate pipeline tag, library name, and base model, and provides links to the paper and the official GitHub repository in the description.

README.md

@@ -0,0 +1,32 @@
+---
+license: apache-2.0
+library_name: peft
+pipeline_tag: text-classification
+base_model: Qwen/Qwen2.5-7B-Instruct
+tags:
+- ai-detection
+- chinese
+- nlpcc
+---
+
+# EnsemJudge (Qwen2.5-7B-Instruct LoRA)
+
+This repository contains a LoRA adapter for **EnsemJudge**, a robust framework for detecting Chinese LLM-generated text. This model is based on Qwen2.5-7B-Instruct and was developed as part of the system that achieved first place in the NLPCC 2025 Shared Task 1.
+
+## Resources
+- **Paper:** [EnsemJudge: Enhancing Reliability in Chinese LLM-Generated Text Detection through Diverse Model Ensembles](https://huggingface.co/papers/2603.27949)
+- **Repository:** [https://github.com/johnsonwangzs/MGT-Mini](https://github.com/johnsonwangzs/MGT-Mini)
+
+## Description
+EnsemJudge addresses the challenge of detecting AI-generated Chinese text, particularly in scenarios involving out-of-domain inputs or adversarial samples. The framework incorporates tailored strategies and ensemble voting mechanisms to demonstrate high effectiveness and reliability.
+
+## Citation
+If you find this work useful, please cite:
+```bibtex
+@article{ensemjudge2025,
+  title={EnsemJudge: Enhancing Reliability in Chinese LLM-Generated Text Detection through Diverse Model Ensembles},
+  author={Guoyu Zhao and others},
+  journal={arXiv preprint arXiv:2603.27949},
+  year={2025}
+}
+```