Arxiv paper - From Bytes to Ideas: Language Modeling with Autoregressive U-Nets | AI Breakdown | Podwise