LLMs can speak in JPEG
Can large language models learn to “speak” image and video file formats? A new study explores an intriguing approach to visual generation using LLMs to model canonical codecs like JPEG and H.264. In this post, we’ll dive deep into how…