TY - RPRT TI - Effective Backdoor Mitigation in Vision-Language Models Depends on the Pre-training Objective AU - Sahil Verma AU - Gantavya Bhatt AU - Avi Schwarzschild AU - Soumye Singhal AU - Arnav Mohanty Das AU - Chirag Shah AU - John P Dickerson AU - Pin-Yu Chen AU - Jeff Bilmes PY - 2025 UR - https://arxiv.org/abs/2311.14948 ID - 2311.14948 ER -