Language Models Are Capable of Metacognitive Monitoring and Control of Their Internal Activations

Open in new window