MEGA Hub

Venus-DefakerOne2: A Vision-Forensics Foundation Model with Broader Coverage

Authors

Do you know Chenfan Qu?You can claim authorship or link another user.Do you know Yachao Liang?You can claim authorship or link another user.Do you know Mingqi Fang?You can claim authorship or link another user.Do you know Jiangwei Xie?You can claim authorship or link another user.Do you know Zhenming Wang?You can claim authorship or link another user.Do you know Baoyue Zhang?You can claim authorship or link another user.Do you know Changjiang Jiang?You can claim authorship or link another user.Do you know Zheng Chang?You can claim authorship or link another user.Do you know Fanli Jin?You can claim authorship or link another user.Do you know Xuekang Zhu?You can claim authorship or link another user.Do you know Kaiwen Feng?You can claim authorship or link another user.Do you know Junchi Li?You can claim authorship or link another user.Do you know Weihong Zhang?You can claim authorship or link another user.Do you know Haoyue Wang?You can claim authorship or link another user.Do you know Chenfeng Zhang?You can claim authorship or link another user.Do you know Yinan Xia?You can claim authorship or link another user.Do you know Fangming Cui?You can claim authorship or link another user.Do you know Dong Fan?You can claim authorship or link another user.Do you know Jingjing Liu?You can claim authorship or link another user.Do you know Tianyi Zhang?You can claim authorship or link another user.Do you know Luobin Chen?You can claim authorship or link another user.Do you know Ji-Zhe Zhou?You can claim authorship or link another user.Do you know Sheng Li?You can claim authorship or link another user.Do you know Wei Wang?You can claim authorship or link another user.Do you know Bin Li?You can claim authorship or link another user.Do you know Ran He?You can claim authorship or link another user.Do you know Xinpeng Zhang?You can claim authorship or link another user.Do you know Linna Zhou?You can claim authorship or link another user.Do you know Jian Liu?You can claim authorship or link another user.Do you know Weiqiang Wang?You can claim authorship or link another user.

Abstract

Real-world forgery is inherently heterogeneous: it manifests not only as digital manipulation but also as physical spoofing, and is carried not only by images but also by videos. Despite substantial progress, a unified anti-forgery foundation model covering such a broad threat spectrum remains absent. In this paper, we present Venus-DefakerOne2, which substantially extends the forensic capability boundary along both the forgery-form and media-type dimensions. Trained on tens of millions of open-source and in-house samples, it incorporates two key designs. Tide-Style-Tuning alternates LoRA and full-parameter fine-tuning to balance computational efficiency, generalization, and hard-sample fitting while mitigating catastrophic forgetting. Entropy-based Test Adaptation determines the optimal prediction modality by analyzing prediction entropy. Venus-DefakerOne2 achieves new state-of-the-art performance across a comprehensive suite of benchmarks, including DeepFake faces (96.7%), AI-generated images (93.3%), tampered documents (87.6%), manipulated natural images (89.6%), physical spoofs (93.1%), and AI-generated videos (96.0%). It significantly outperforms both specialized expert models and its predecessor, DefakerOne, establishing itself as a powerful and versatile foundation model for binary-decision forensic tasks.

Community

00