MT Forcing: Towards Stable and Consistent Multi-Turn Image Editing with Autoregressive Diffusion

A research paper titled "MT Forcing: Towards Stable and Consistent Multi-Turn Image Editing with Autoregressive Diffusion" proposes an approach to reduce drift in multi-turn image editing, where existing methods degrade after only a few editing turns. The work targets two failure modes: unstable visual quality and inconsistency across successive edits, aiming to make interactive visual content creation more reliable for users.

Multi-turn image editing aims to modify an image according to a sequence of instructions, enabling more effective interactive visual content creation for users. However, existing methods often exhibit noticeable drift after only a few editing turns, manifesting in two aspects: 1 unstable visual qua