Add model card for REVEAL

This PR adds a comprehensive model card for REVEAL. It includes:
- Links to the paper: [Reasoning-Aware AIGC Detection via Alignment and Reinforcement](https://huggingface.co/papers/2604.19172).
- Link to the project page: https://aka.ms/reveal.
- Link to the Github repository: https://github.com/microsoft/AnthropomorphicIntelligence.
- Relevant metadata for `library_name: transformers` and `pipeline_tag: text-classification`.

README.md

@@ -0,0 +1,25 @@
+---
+library_name: transformers
+pipeline_tag: text-classification
+---
+
+# REVEAL: Reasoning-Aware AIGC Detection via Alignment and Reinforcement
+
+REVEAL is a detection framework designed to identify AI-generated content (AIGC) by generating interpretable reasoning chains before final classification. This approach enhances both the accuracy and the transparency of the detection process.
+
+- **Paper:** [Reasoning-Aware AIGC Detection via Alignment and Reinforcement](https://huggingface.co/papers/2604.19172)
+- **Project Page:** [https://aka.ms/reveal](https://aka.ms/reveal)
+- **Repository:** [https://github.com/microsoft/AnthropomorphicIntelligence](https://github.com/microsoft/AnthropomorphicIntelligence)
+
+## Introduction
+
+The REVEAL framework addresses the challenges of reliable AIGC detection as Large Language Models (LLMs) continue to evolve. It utilizes a two-stage training strategy:
+1. **Supervised Fine-tuning (SFT):** Establishes the model's reasoning capabilities.
+2. **Reinforcement Learning (RL):** Refines accuracy and logical consistency while reducing hallucinations in the reasoning process.
+
+The model is trained on **AIGC-text-bank**, a comprehensive multi-domain dataset encompassing diverse LLM sources and authorship scenarios.
+
+## Model Details
+- **Architecture**: Based on the `qwen3` model type.
+- **Task**: Text Classification (AIGC Detection).
+- **Training**: Fine-tuned on the AIGC-text-bank dataset using reasoning-aware alignment and reinforcement learning.