论文

PaTh:把递归推理引入扩散模型,像素空间解 Sudoku

Think Before You Paint: Recursive Latent Reasoning for Diffusion Models

精选理由

一篇挺有意思的论文:给扩散模型外挂一个小推理网络,10M 参数就把最难 Sudoku 从 4.1% 拉到 71.2%,不用符号监督。

PaTh(Painter-Thinker)是一个让扩散模型具备推理能力的框架:一个 10M 参数的小型递归网络 Thinker 在每个去噪步骤内对编码噪声图像的 token 网格做潜在状态细化,并通过 ControlNet 适配器驱动冻结的扩散模型 Painter。训练只用标准重建损失,不需要符号目标、求解器或验证器。PaTh 在困难 MNIST Sudoku 上达到 92.5% 的解决率(此前最佳 75%),极端难度下达到 71.2%(此前最佳 4.1%),参数量 10M 对比标准扩散模型的 82M。在迷宫、Queens 和 CLEVR 空间关系任务上也有提升,且问题规模越大优势越明显,还能修复扩散模型无法自行纠正的错误。

原文 · arXiv cs.AI

Think Before You Paint: Recursive Latent Reasoning for Diffusion Models

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. When a discrete symbolic representation is available, recursive methods such as the Tiny Recursive Model (TRM) solve even hard instances of these puzzles. We ask how such reasoning can be carried over to pixels, where no symbolic representation is available. 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. The Thinker is trained with the standard reconstruction loss alone, without symbolic targets, a solver, or a verifier. PaTh solves 92.5% of hard MNIST Sudoku puzzles (prior best 75%) and 71.2% of extreme ones (prior best 4.1%), with 10M parameters against 82M for a standard diffusion model. It also improves on mazes, Queens, and CLEVR scenes with specified spatial relations, and its advantage grows with problem size. Diagnostic experiments show that PaTh recovers from injected mistakes that the diffusion model cannot repair, especially when many cells are wrong. 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.