Zobrazeno 1 - 8
of 8
pro vyhledávání: '"Nie, Yuzhou."'
Existing works have established multiple benchmarks to highlight the security risks associated with Code GenAI. These risks are primarily reflected in two areas: a model potential to generate insecure code (insecure coding) and its utility in cyberat
Externí odkaz:
http://arxiv.org/abs/2410.11096
Modern large language model (LLM) developers typically conduct a safety alignment to prevent an LLM from generating unethical or harmful content. Recent studies have discovered that the safety alignment of LLMs can be bypassed by jailbreaking prompts
Externí odkaz:
http://arxiv.org/abs/2406.08725
Recent studies developed jailbreaking attacks, which construct jailbreaking prompts to fool LLMs into responding to harmful questions. Early-stage jailbreaking attacks require access to model internals or significant human efforts. More advanced atta
Externí odkaz:
http://arxiv.org/abs/2406.08705
Autor:
Nie, Yuzhou., Wang, Yanting., Jia, Jinyuan., De Lucia, Michael J., Bastian, Nathaniel D., Guo, Wenbo., Song, Dawn.
One key challenge in backdoor attacks against large foundation models is the resource limits. Backdoor attacks usually require retraining the target model, which is impractical for very large foundation models. Existing backdoor attacks are mainly de
Externí odkaz:
http://arxiv.org/abs/2405.16783
Akademický článek
Tento výsledek nelze pro nepřihlášené uživatele zobrazit.
K zobrazení výsledku je třeba se přihlásit.
K zobrazení výsledku je třeba se přihlásit.
Autor:
Wang, Wenwen, Zhang, Xiaochen, Gui, Ping, Zou, Qizhen, Nie, Yuzhou, Ma, Shenglin, Zhang, Shirong
Publikováno v:
BMC Cancer; 9/5/2024, Vol. 24 Issue 1, p1-20, 20p
Publikováno v:
Proceedings of SPIE; 5/24/2022, Vol. 12176, p1217613-1217613, 1p
Publikováno v:
Proceedings of SPIE; May 2022, Vol. 12176 Issue: 1 p1217613-1217613-14