Skip to content

Translation in progress Some entries still include text in the original language.

NVIDIA 技术博客:可信 AI 与网络安全· Joseph Lucas·Indexed by this site · Originally published

NVIDIA 如何通过结构化应用保护 KV cache

Structuring Applications to Secure the KV Cache

AI summary

NVIDIA 提出通过结构化应用设计来保护 KV cache。输入结构会影响 Transformer 模型(如 LLM、VLM)的输出,而实际提示词往往不止是简单的用户提问,常由系统指令、上下文数据和用户输入等来源动态拼装而成。

Read originaldeveloper.nvidia.com