Vol. 2 No. 4 (2025): 中文大語言模型越獄框架
如何突破大語言模型的安全防護機制?
採用場景僞裝手段,將惡意指令隱匿於安全語境之中,並通過指令拆分技術將風險內容碎片化,從而利用模型的推理能力重組並執行有效載荷。
已發表:
2025-12-31
lssue 4, Vol. 2 (2025) will be published on December 31,2025. The organizing institution will be changed to "HongKong Turing General AI Research Institute CO., Limited", while the company website, office address, contact information, submission requirements, and other details remain unchanged.