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

 (github.com)

1 points | by torsi0n 6 hours ago ago

No comments yet.