README_EN.md

metadata

license: mit
datasets:
  - anyangsong/Blockchain-Sensitive-Detect-Data
language:
  - zh
  - en
metrics:
  - accuracy
  - f1
  - precision
  - recall

Blockchain-Sensitive-Detect-Checkpoints

中文文档

Checkpoints for the model used in the Blockchain Sensitive Information Detection project at Children's Hospital of Fudan University.

Repository: https://github.com/anyangsong/Blockchain-Sensitive-Detect

Dataset: https://huggingface.co/datasets/anyangsong/Blockchain-Sensitive-Detect-Data

Checkpoints: https://huggingface.co/anyangsong/Blockchain-Sensitive-Detect-Checkpoints

Directory Structure

.
├── machine_video_detectors/
│   └── checkpoints/
├── sensitive_image_detector/
│   └── weights/
└── video_violent_detector/
    ├── ckpt/
    └── data/
        ├── FlowTest/
        ├── RGBTest/
        └── vggish-features/

Instructions

This repository open-sources the model checkpoints used for detecting sensitive videos, unknown sensitive images, and machine-generated videos in the project.

The sensitive text, image, and audio detection components of this project mainly rely on prompt engineering and complex reasoning. They previously used Qwen2.5-VL-7B-Instruct and Qwen2-Audio-7B-Instruct, respectively.

You can directly use the original parameters of these two large language models, or fine-tune them using standard RL/agentic RL frameworks.

1. Move the Qwen2.5-VL-7B-Instruct and Qwen2-Audio-7B-Instruct weight directories to /remote-home/sensitive-detect/checkpoints/.
2. Move the video_violent_detector directory to /remote-home/sensitive-detect/detectors/video_violent_detector/data/.
3. Move the sensitive_image_detector directory to /remote-home/sensitive-detect/detectors/sensitive_image_detector/weights/.
4. Move the machine_video_detector directory to /remote-home/sensitive-detect/detectors/machine_video_detector/checkpoints/.

Citation

If you find this project and its dataset useful for your research and applications, please cite using the Bibtex:

@misc{anyang_song_2026,
    author       = { Anyang Song and Changkai Ji and Hao Fang and Yufeng Liu and Yang Li and Junjie He and Ying Cheng and Ruiwei Zhao and Yuejie Zhang and Rui Feng },
    title        = { Blockchain-Sensitive-Detect },
    year         = 2026,
    url          = { https://huggingface.co/datasets/anyangsong/Blockchain-Sensitive-Detect-Data },
    doi          = { 10.57967/hf/10293 },
    publisher    = { Hugging Face }
}