AF - Connecting the Dots: LLMs can Infer & Verbalize Latent Structure from Training Data by Johannes Treutlein | The Nonlinear Library | Podwise