Initial dataset card for MENTOR dataset

README.md

@@ -0,0 +1,58 @@
+---
+task_categories:
+- text-to-image
+tags:
+- multimodal
+- autoregressive
+- image-generation
+license: mit
+---
+
+# MENTOR Dataset
+
+This repository contains dataset components used in the paper "[MENTOR: Efficient Multimodal-Conditioned Tuning for Autoregressive Vision Generation Models](https://huggingface.co/papers/2507.09574)".
+
+MENTOR (Multimodal-conditioned Tuning for Autoregressive Vision Generation Models) proposes a novel autoregressive (AR) framework designed for efficient multimodal image generation. It combines an AR image generator with a two-stage training paradigm to achieve fine-grained, token-level alignment between multimodal inputs and image outputs.
+
+*   **Project Page**: [https://haozhezhao.github.io/MENTOR.page](https://haozhezhao.github.io/MENTOR.page)
+*   **Code**: [https://github.com/HaozheZhao/MENTOR](https://github.com/HaozheZhao/MENTOR)
+
+## Dataset Components
+
+The MENTOR framework relies on a two-stage training process, each utilizing a specific dataset component:
+
+1.  **Stage 1: Multimodal Alignment Dataset**: This dataset is used in the first training stage to establish robust pixel- and semantic-level alignment. Tasks covered include image reconstruction, object segmentation, and text-to-image generation.
+2.  **Stage 2: Multimodal Instruction Tuning Dataset**: This dataset is used in the second training stage to balance the integration of multimodal inputs and enhance generation controllability.
+
+## Sample Usage
+
+You can download the MENTOR dataset components (Stage 1 and Stage 2) using the Hugging Face CLI:
+
+```bash
+# Download Stage-1 dataset
+huggingface-cli download BleachNick/Mentor_Stage1 --repo-type dataset --local-dir Mentor_Stage1
+cd Mentor_Stage1
+cat stage1_data.tar.gz.part-* | pv | tar -xzf -
+cd ..
+
+# Download Stage-2 dataset
+huggingface-cli download BleachNick/Mentor_Stage2 --repo-type dataset --local-dir Mentor_Stage2
+cd Mentor_Stage2
+cat stage2_data.tar.gz.part-* | pv | tar -xzf -
+cd ..
+```
+
+For more detailed instructions on training and model usage, please refer to the [official GitHub repository](https://github.com/HaozheZhao/MENTOR).
+
+## Citation
+
+If you find MENTOR and its associated datasets useful for your research, please cite the paper:
+
+```bibtex
+@inproceedings{zhao2024mentor,
+  title={MENTOR: Efficient Multimodal-Conditioned Tuning for Autoregressive Vision Generation Models},
+  author={Zhao, Haozhe* and Cai, Zefan* and Si, Shuzheng and Chen, Liang and
+          Gu, Jiuxiang and Xiao, Wen and Hu, Junjie},
+  year={2024}
+}
+```