研究人员开发了一种新颖的方法，通过利用多模态大型语言模型（MLLMs）来训练深度伪造检测器。该方法使用语言作为正则化机制，以提高检测器的泛化能力和可解释性。该系统采用双编码器架构和两阶段训练过程，包括强化学习以鼓励分类前的描述性推理，从而显著提高性能并提供可解释的输出。 AI
影响 通过提高泛化能力和可解释性来增强深度伪造检测能力，这对于打击人工智能生成的虚假信息至关重要。
排序理由 该集群包含一篇详细介绍深度伪造检测新研究方法的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
arXiv:2605.31192v1 Announce Type: new Abstract: Recently, thanks to the advent of Multimodal-LLMs, deepfake detectors are striving not only to be generalizable but also interpretable. We propose that these two challenges can effectively be tackled jointly, since describable artif…
Recently, thanks to the advent of Multimodal-LLMs, deepfake detectors are striving not only to be generalizable but also interpretable. We propose that these two challenges can effectively be tackled jointly, since describable artifacts typically generalize better, opening the po…