Connecting the Dots: LLMs can Infer and Verbalize Latent Structure from Disparate Training Data 1

Open in new window