Ant Group's Venus-DefakerOne2 hits 96% accuracy detecting AI-generated videos
inclusionAI · hf · 2026-11-02
GuangJian Team at Ant Group, with CSIG's Digital Media Forensics committee, released Venus-DefakerOne2, a unified vision-forensics foundation model extending detection across forgery forms (digital manipulation and physical spoofing) and media types (images and videos).
Key designs
- Tide-Style-Tuning: alternates LoRA and full-parameter fine-tuning to balance efficiency, generalization, and hard-sample fitting while mitigating catastrophic forgetting
- Entropy-based Test Adaptation: picks the optimal prediction modality by analyzing prediction entropy
Trained on tens of millions of open-source and in-house samples, it sets new SOTA across benchmarks: 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%), significantly outperforming specialized expert models and its predecessor DefakerOne.
More from Safety
- Auditing my MCP credential broker: the worst hole was the config file — Relevant_Mousse_9117 · 2026-11-02
- Cambricon co-founder urges shared US-China AI ethics standards for leading models — pstAsiatech · 2026-11-02
- Vitalik's "post-AI cryptography": pseudonyms are dead, per-action anonymity is in — devanshmehta · 2026-11-02
- 6 prompts to see what ChatGPT quietly knows about you — and wipe it — Aiden_Tech_Ai · 2026-11-02
- European supermarket freezers hit by second suspected cyber-attack — fergie · 2026-11-02
- Insider: critical infrastructure firm flunked security audit, still delays basic fixes — thedealdirector · 2026-11-02