특정 트리거가 발생할 때만 기만적인 행동을 보이는 LLMs를 훈련하는 방법 like Spy even after AI Alignment (safety learning)
Sleeper Agents Attack
Creator
Creator
Seonglae ChoCreated
Created
2024 Jan 15 17:3Editor
Editor
Seonglae ChoEdited
Edited
2024 Jan 16 14:51Refs
Seonglae ChoSeonglae Cho
