Imperio: Language-Guided Backdoor Attacks for Arbitrary Model Control

Open in new window