1 paper
Bowen Sun, Chaozhuo Li, Yaodong Yang +2
Decompositional jailbreaks pose a critical threat to large language models (LLMs) by allowing adversaries to fragment a malicious objective into a sequence of individually benign q…