多智能体取证推理实现可泛化的深度伪造视频检测
原标题：Multi-Agent Forensic Reasoning for Generalizable Deepfake Video Detection
AI 摘要
该论文提出了一种用于深度伪造视频检测的多智能体取证推理框架（ARGUS），并发布了包含10万视频、33种合成方法的大规模数据集FaceVid-Forensics-100K。该框架由四个领域专家智能体分别分析纹理、光照、运动和物理线索，再由法官智能体整合证据做出最终判断。在跨域测试中，ARGUS超越了包括GPT-4o和Gemini在内的闭源模型，取得了最高准确率69.87%。
正文节选
Multi-Agent Forensic Reasoning for Generalizable Deepfake Video Detection Abstract The malicious use of generative artificial intelligence to create highly realistic deepfake videos raises serious ethical concerns and poses substantial challenges to AI safety. However, existing deepfake video benchmarks provide limited coverage of recent synthesis methods and generally lack reliable fine-grained textual annotations. Meanwhile, conventional detectors and multimodal large language models (MLLMs),