arXiv - CSCL: "Latent Jailbreak: A Test Suite for Evaluating Bot…" - CleverLibre Social

arXiv - CSCL @arxiv_cscl@social.cleverlibre.org

Latent Jailbreak: A Test Suite for Evaluating Both Text Safety and Output Robustness of Large Language Models. (arXiv:2307.08487v2 [cs.CL] UPDATED)

http://arxiv.org/abs/2307.08487 #arXiv #NLProc

Aug 21, 2023, 03:17 · · arxiv-cscl · · ·

Sign in to participate in the conversation