AI RESEARCH
Autoregressive Visual Generation Needs a Prologue
arXiv CS.LG
•
ArXi:2605.06137v1 Announce Type: cross In this work, we propose Prologue, an approach to bridging the reconstruction-generation gap in autoregressive (AR) image generation. Instead of modifying visual tokens to satisfy both reconstruction and generation, Prologue generates a small set of prologue tokens prepended to the visual token sequence. These prologue tokens are trained exclusively with the AR cross-entropy (CE) loss, while visual tokens remain dedicated to reconstruction.