Reading a Model’s Mind with the Jacobian Lens
中文摘要
Anthropic 发现了一种利用 Jacobian 镜头的方法，能在语言模型输出前读取其内部思维过程。
English Summary
Anthropic found a way to use the Jacobian lens to see a language model's internal thoughts before it produces any output.
Original Excerpt
Anthropic found a way to see what a language model is thinking before it says anything. Continue reading on Medium »