Abstract

In this work, we introduce Janus-Pro, an advanced version of the previouswork Janus. Specifically, Janus-Pro incorporates (1) an optimized trainingstrategy, (2) expanded training data, and (3) scaling to larger model size.With these improvements, Janus-Pro achieves significant advancements in bothmultimodal understanding and text-to-image instruction-following capabilities,while also enhancing the stability of text-to-image generation. We hope thiswork will inspire further exploration in the field. Code and models arepublicly available.

 

Quick Read (beta)