Think in Strokes, Not Pixels: Process-Driven Image Generation via Interleaved Reasoning
This addresses the challenge of making image generation more human-like and interpretable for AI researchers and practitioners, though it appears incremental in building on existing multimodal models.
The paper tackles the problem of generating images through a multi-step reasoning process rather than single-step synthesis, introducing a method that decomposes generation into interleaved textual planning and visual refinement stages. The result is a process-driven approach that achieves competitive performance on text-to-image benchmarks, with explicit and interpretable intermediate states.
Humans paint images incrementally: they plan a global layout, sketch a coarse draft, inspect, and refine details, and most importantly, each step is grounded in the evolving visual states. However, can unified multimodal models trained on text-image interleaved datasets also imagine the chain of intermediate states? In this paper, we introduce process-driven image generation, a multi-step paradigm that decomposes synthesis into an interleaved reasoning trajectory of thoughts and actions. Rather than generating images in a single step, our approach unfolds across multiple iterations, each consisting of 4 stages: textual planning, visual drafting, textual reflection, and visual refinement. The textual reasoning explicitly conditions how the visual state should evolve, while the generated visual intermediate in turn constrains and grounds the next round of textual reasoning. A core challenge of process-driven generation stems from the ambiguity of intermediate states: how can models evaluate each partially-complete image? We address this through dense, step-wise supervision that maintains two complementary constraints: for the visual intermediate states, we enforce the spatial and semantic consistency; for the textual intermediate states, we preserve the prior visual knowledge while enabling the model to identify and correct prompt-violating elements. This makes the generation process explicit, interpretable, and directly supervisable. To validate proposed method, we conduct experiments under various text-to-image generation benchmarks.