The "Canary Token" Trick: How to Know When an AI is Ignoring Your System Instructions
中文摘要
学习利用“金丝雀令牌”检测 AI 是否忽略系统指令，从而识别指令漂移或提示词注入。
English Summary
Learn how to use "Canary Tokens" to detect if an AI is ignoring system instructions, helping identify instruction drift or prompt injection.
原文节选
Continue reading on Medium »