BAIT: Large Language Model Backdoor Scanning by Inverting Attack Target

Published in 2025 IEEE Symposium on Security and Privacy (S&P 2025), 2025