README.md

---
license: cc-by-nc-sa-4.0
task_categories:
- visual-question-answering
language:
- en
size_categories:
- 1K<n<10K
---


# ADOPD-Dataset-6K

<p align="center">
  📄 <a href="https://arxiv.org/abs/2608.09789">Paper</a> &nbsp;|&nbsp;
  💻 <a href="https://github.com/withTai/ADOPD">Code</a> &nbsp;|&nbsp;
  🤗 <a href="https://huggingface.co/withstaticTai/ADOPD">Model</a> &nbsp;|&nbsp;
</p>


## Introdution

**ADOPD-Dataset-6K** is the released training dataset in paper [ADOPD: Reference-Privileged On-Policy Distillation for MLLM-Based Industrial Anomaly Detection](https://arxiv.org/abs/2608.09789).

ADOPD-6K is a 6,000-record paired visual anomaly reasoning training set derived from Real-IAD. Each pair consists of one anomalous image and its corresponding normal counterfactual image. The paired dataset is used to generate matched and mismatched privileged references for the teacher during training. The student receives the original 1024 × 1024 inspection image without a defect bounding box. In contrast, teacher supervision uses a localized crop in which the target region is indicated by a red bounding box.


## Resource

ADOPD-6K is derived from the Real-IAD dataset.
```
- Real-IAD project: https://realiad4ad.github.io/Real-IAD/
- Real-IAD dataset page: https://huggingface.co/datasets/Real-IAD/Real-IAD
```


## Citation

```bibtex
@article{he2026adopd,
  title={ADOPD: Reference-Privileged On-Policy Distillation for MLLM-Based Industrial Anomaly Detection},
  author={He, Jingtai and Meng, Shiyuan and Meng, Wenchao and Yang, Qinmin},
  journal={arXiv preprint arXiv:2608.09789},
  year={2026}
}
```

## License

Released under CC BY-NC-SA 4.0, same as the original data Real-IAD.