#survey

The Sentinel's Dilemma: Guarding AI from Hidden Threats

A survey of backdoor attacks on large language models — how hidden triggers are implanted during training, and the detection and defense strategies emerging against them.