We can still sleep peacefully — or so we thought. Steganography via internal activations is already possible in small language models — a potential first step toward persistent hidden reasoning.

Introduction
In this post we continue to work on the project, started in our previous  post, of exploring  hidden reasoning of large language models…