MM-MergeBench / README.md
AuroraZengfh's picture
Update README.md
c05d571 verified
# Card of Dataset for Multimodal Large Language Model
## Dataset details and sources
This dataset is constructed using publicly available instruction tuning datasets, including COCO, ScienceQA, VizWiz, ImageNet, VQAv2, ImageNet-R, Flickr30k, OCRVQA, Screen2words and TabMWP.
### Seen datasets for merging
| Dataset | Image Source | Download Path |
| :----: | :----: | :----: |
| ScienceQA | ScienceQA | [images](https://drive.google.com/drive/folders/1w8imCXWYn2LxajmGeGH_g5DaL2rabHev) |
| VizWiz | VizWiz | [images](https://opendatalab.org.cn/OpenDataLab/VizWiz-Captions) |
| ImageNet | ImageNet | [images](https://image-net.org/challenges/LSVRC/index.php) |
| VQAv2, Flickr30k | COCO2014 | [images](http://images.cocodataset.org/zips/train2014.zip) |
| IconQA | IconQA | [images](https://iconqa2021.s3.us-west-1.amazonaws.com/iconqa_data.zip) |
| Flickr30k | Flickr30k | [images](https://github.com/BryanPlummer/flickr30k_entities) |
| OCRVQA | OCRVQA | [images](https://drive.google.com/drive/folders/1_GYPY5UkUy7HIcR0zq3ZCFgeZN7BAfm_) |
### Unseen datasets for merging
| Dataset | Image Source | Download Path |
| :----: | :----: | :----: |
| AOKVQA | COCO2014 | [images](http://images.cocodataset.org/zips/val2014.zip) |
| ImageNet-R | ImageNet-R | [images](https://people.eecs.berkeley.edu/~hendrycks/imagenet-r.tar) |
| Screen2words | Screen2words | [images](https://huggingface.co/datasets/pinkmooncake/rico-screen2words) |
| TabMWP | TabMWP | [images](https://github.com/lupantech/PromptPG/tree/main/data/tabmwp)|
Please download these datasets before proceeding with instruction tuning.
-----------------------------------------------------------------------------------------------------------------
## **RobustMerge**
Paper: [RobustMerge: Parameter-Efficient Model Merging for MLLMs with Direction Robustness](https://arxiv.org/abs/2502.17159)
Code: [github.com/AuroraZengfh/RobustMerge](https://github.com/AuroraZengfh/RobustMerge)
If you find this work useful, consider citing our paper as follows:
```bibtex
@article{zeng2025parameter,
title={RobustMerge: Parameter-Efficient Model Merging for MLLMs with Direction Robustness},
author={Zeng, Fanhu and Guo, Haiyang and Zhu, Fei and Shen, Li and Tang, Hao},
journal={arXiv preprint arXiv:2502.17159},
year={2025}
}
```
---
license: mit
---