In this work, we introduce:
📍Perception Pretext RL: an algorithm that leverages simple perception pretext tasks to elevate detection performance. Compatible with R1-paradigm frameworks.
📍VideoVeritas Model: a framework that integrates fine-grained perception and fact-based reasoning for AI-generated video detection.
📍MintVid Dataset: a light yet high-quality AI-generated video dataset that involves three parts: (1) general content, (2) facial, and (3) fact-based videos.
The growing capability of video generation poses escalating security risks, making reliable detection increasingly essential. In this paper, we introduce VideoVeritas, a framework that integrates fine-grained perception and fact-based reasoning. We observe that while current multi-modal large language models (MLLMs) exhibit strong reasoning capacity, their granular perception ability remains limited. To mitigate this, we introduce Joint Preference Alignment and Perception Pretext Reinforcement Learning (PPRL). Specifically, rather than directly optimizing for detection task, we adopt general spatiotemporal grounding and self-supervised object counting in the RL stage, enhancing detection performance with simple perception pretext tasks. To facilitate robust evaluation, we further introduce MintVid, a light yet high-quality dataset containing 3K videos from 9 state-of-the-art generators, along with a real-world collected subset that has factual errors in content. Experimental results demonstrate that existing methods tend to bias towards either superficial reasoning or mechanical analysis, while VideoVeritas achieves more balanced performance across diverse benchmarks.
conda create -n videoveritas python=3.10
conda activate videoveritas
# Install the dependencies
pip install -e .Download VideoVeritas 🔥🔥🔥. We recommend using vLLM for model deployment:
sh self_scripts/deploy/deploy_model.sh /path/to/your/modelInference on a single video:
python self_scripts/infer/infer_vllm_single.py \
--video_path /path/to/your/videoDownload the MintVid dataset.
Change the json file path in ./swift/llm/dataset/dataset/data_utils.py and the video path in the json files.
sh self_scripts/infer/infer_mintvid.shIf you find our work useful, please cite our paper:
@article{tan2026videoveritas,
title={VideoVeritas: AI-Generated Video Detection via Perception Pretext Reinforcement Learning},
author={Tan, Hao and Lan, Jun and Shi, Senyuan and Tan, Zichang and Yu, Zijian and Zhu, Huijia and Wang, Weiqiang and Wan, Jun and Lei, Zhen},
journal={arXiv preprint arXiv:2602.08828},
year={2026}
}
This repo is released under the Apache 2.0 License.
This repo benefits from ms-swift and DeepfakeBench. Thanks for their great works!