arXiv - CSCL: "Red teaming ChatGPT via Jailbreaking: Bias, Robus…" - Qoto Mastodon

arXiv - CSCL @arxiv_cscl@qoto.org

Red teaming ChatGPT via Jailbreaking: Bias, Robustness, Reliability and Toxicity. (arXiv:2301.12867v4 [cs.CL] UPDATED)

http://arxiv.org/abs/2301.12867 #arXiv #NLProc

May 30, 2023, 03:06 · · arxiv-cscl · · ·

Sign in to participate in the conversation