Translation in progress Some entries still include text in the original language.
Oct 13 · Tuesday
Agent safety · Papers
Found 2 papers提出移动 Agent 轨迹审计器 MATE,判断轨迹是否违反可编辑安全策略
- Published
- Evaluated models
- MATE-0.5B, MATE-1.5B, MATE-3B
AbstractMATE 用合成轨迹微调轻量审计器,在可编辑策略下判断移动智能体轨迹是否违规并给出解释。
MATE 是面向移动 GUI 智能体的轻量策略条件轨迹审计器,在给定自然语言安全策略时判断轨迹是否违规并解释原因。
提出 HazardAuditor,用可执行轨迹训练计算机使用 Agent 的运行时守卫
- Published
- Evaluated models
- HazardAuditor, HazardAuditor-SFT
You have reached the end