GLID: Pre-training a Generalist Encoder-Decoder Vision Model | Read Paper on Bytez