Think Before You Paint: Recursive Latent Reasoning for Diffusion Models

Key points

• Diffusion models generate realistic images but often fail on visual reasoning tasks, such as filling in a Sudoku or drawing the path through a maze.1
• When a discrete symbolic representation is available, recursive methods such as the Tiny Recursive Model (TRM) solve even hard instances of these puzzles.1
• We ask how such reasoning can be carried over to pixels, where no symbolic representation is available.1
• Together, these results show that reasoning mechanisms developed for symbolic data can be integrated into pixel-space diffusion without symbolic supervision, opening a path toward generating data under increasingly complex constraints.1

Sources (1)

• [1]Think Before You Paint: Recursive Latent Reasoning for Diffusion Models
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 7, 11:36 AM

  We propose Painter-Thinker (PaTh): a small recursive network (the Thinker) reasons over a grid of learned tokens that encode the noisy image and the conditioning, refines a latent state within every denoising step, and steers a frozen diffusion model (the Painter) through ControlNet adapters.

  Diffusion models generate realistic images but often fail on visual reasoning tasks, such as filling in a Sudoku or drawing the path through a maze.

Extractive summary: sentences quoted from the sources.