ChristinaW/MoDE-official

MoDE Dataset

This repository accompanies the NeurIPS 2025 paper:
“Mitigating Intra- and Inter-modal Forgetting in Continual Learning of Unified Multimodal Models.”
📄 arXiv • 💻 GitHub

---

Overview

The MoDE dataset is adapted from the CoIN benchmark.

The repository includes:

• data/ — Processed training and validation splits for all tasks
• instructions/ — Evaluation instructions and prompts

---

Datasets Download

Please refer to the original CoIN dataset documentation to download the raw datasets:

🔗 https://github.com/zackschen/CoIN?tab=readme-ov-file#dataset

The tasks include:

• ScienceQA
• ImageNet
• TextVQA
• GQA
• VizWiz

---

Citation

If you use MoDE or its processed data in your research, please cite the associated NeurIPS 2025 paper.

@inproceedings{weimitigating,
  title={Mitigating Intra-and Inter-modal Forgetting in Continual Learning of Unified Multimodal Models},
  author={Wei, Xiwen and Munir, Mustafa and Marculescu, Radu},
  booktitle={The Thirty-ninth Annual Conference on Neural Information Processing Systems}
}