Reformulates diffusion sampling as a graph-theoretic planning problem that dynamically allocates compute to the most difficult denoising stages.
March 17, 2026
Original Paper
Chain-of-Trajectories: Unlocking the Intrinsic Generative Optimality of Diffusion Models via Graph-Theoretic Planning
arXiv · 2603.14704
The Takeaway
Implements a 'System 2' deliberative sampling schedule that is content-aware and train-free. It significantly improves generation stability and quality by identifying and spending more time on the 'DNA' (difficulty signature) of a specific image path.
From the abstract
Diffusion models operate in a reflexive System 1 mode, constrained by a fixed, content-agnostic sampling schedule. This rigidity arises from the curse of state dimensionality, where the combinatorial explosion of possible states in the high-dimensional noise manifold renders explicit trajectory planning intractable and leads to systematic computational misallocation. To address this, we introduce Chain-of-Trajectories (CoTj), a train-free framework enabling System 2 deliberative planning. Centra