README.md

---
language:
- en
license: mit
size_categories:
- 1M<n<10M
task_categories:
- visual-question-answering
- image-text-to-text
pretty_name: ABC-Pretraining-Data
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: caption
    dtype: string
  - name: url
    dtype: string
  - name: id
    dtype: int64
  - name: image
    dtype: string
  - name: negatives
    sequence: int64
  splits:
  - name: train
    num_bytes: 2289772991
    num_examples: 2252041
  download_size: 1855548818
  dataset_size: 2289772991
tags:
- visual
---

## ABC Pretraining Data

<!-- Provide a quick summary of the dataset. -->
This the the pretraining data for ABC. This dataset is derived from Google's [Conceptual Captions](https://ai.google.com/research/ConceptualCaptions/) dataset.
The each item in the dataset contain a URL where the corresponding image can be downloaded and mined negatives for each item. Full dataaset is ~300 GB of images. For a detailed description of how we mined the negatives please check out our ppaer ;).   
**Update** I have added the images to this repository, for an example of how to use and download this dataset see our [repository](https://github.com/TIGER-AI-Lab/ABC).

## Paper and Website

For more information, please refer to [Website](https://tiger-ai-lab.github.io/ABC/).

## Citation

If you find any of our work helpful please connsider citing:

```
@misc{schneider2025abcachievingbettercontrol,
      title={ABC: Achieving Better Control of Multimodal Embeddings using VLMs}, 
      author={Benjamin Schneider and Florian Kerschbaum and Wenhu Chen},
      year={2025},
      eprint={2503.00329},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2503.00329}, 
}
```