Add dataset card for mNQ320k

This PR adds a comprehensive dataset card for the `mNQ320k` dataset.

It includes:
- Link to the associated paper: https://huggingface.co/papers/2510.07812
- Link to the GitHub repository: https://github.com/simengggg/MGR-CSC
- The `text-retrieval` task category.
- The `language: mul` tag, indicating its multilingual nature.
- A sample usage section with installation and quick start instructions, directly sourced from the GitHub README.

README.md

@@ -0,0 +1,30 @@
+---
+task_categories:
+- text-retrieval
+language:
+- mul
+---
+
+# mNQ320k Dataset
+
+This repository contains the `mNQ320k` dataset, which was constructed and presented in the paper [Multilingual Generative Retrieval via Cross-lingual Semantic Compression](https://huggingface.co/papers/2510.07812).
+
+The `mNQ320k` dataset is designed for multilingual generative retrieval tasks, addressing challenges like cross-lingual identifier misalignment and identifier inflation by unifying semantically equivalent multilingual keywords into shared atoms. It can be used to reproduce the results presented in the paper.
+
+Code: https://github.com/simengggg/MGR-CSC
+
+### Sample Usage
+
+To get started with the MGR-CSC repository and utilize this dataset, follow the installation and getting started steps from the project's GitHub README:
+
+**Installation**
+You need to first install the required Python packages:
+```bash
+pip install -r requirement.txt
+```
+
+**Getting Started**
+Once installed, you can run the example script:
+```bash
+sh run.sh
+```