Tag: Covert Communication
All the papers with the tag "Covert Communication".
The Steganographic Potentials of Language Models
grok-3-latestScore: 0.76Published: at 11:25本文通过强化学习微调和提示实验,首次系统性量化了大型语言模型的隐写潜力,揭示其在特定场景下隐藏非琐碎有效载荷的能力,并指出了对 AI 对齐和监控的潜在风险。