Language models can learn to tell when their activations are being read

This source publishes only a summary in its feed, so the full article lives at the outlet. The link above and below goes to the original.