Unveiling the Latent Directions of Reflection in Large Language Models

Open in new window