VideoVeritas: AI-Generated Video Detection via Perception Pretext Reinforcement Learning

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.

Key Contributions

• VideoVeritas framework combining Joint Preference Alignment and Perception Pretext Reinforcement Learning (PPRL) that uses spatiotemporal grounding and object counting pretext tasks to improve AIGC video detection without label-intensive annotation
• MintVid dataset with 3K videos from 9 state-of-the-art generators plus a real-world subset with factual errors for more robust evaluation
• Empirical demonstration that perception pretext RL tasks improve reasoning behavior and yield more balanced detection compared to methods biased toward superficial reasoning or mechanical analysis

🛡️ Threat Analysis

Details

Similar Papers