WiT: Waypoint Diffusion Transformers via Trajectory Conflict Navigation
This addresses a specific bottleneck in pixel-space diffusion models for image generation, offering an incremental improvement over existing methods.
The paper tackles the problem of trajectory conflicts in pixel-space Flow Matching models by proposing Waypoint Diffusion Transformers (WiT), which factorizes the vector field via semantic waypoints to disentangle generation trajectories. On ImageNet 256x256, WiT beats pixel-space baselines and accelerates training convergence by 2.2x.
While recent Flow Matching models avoid the reconstruction bottlenecks of latent autoencoders by operating directly in pixel space, the lack of semantic continuity in the pixel manifold severely intertwines optimal transport paths. This induces severe trajectory conflicts near intersections, yielding sub-optimal solutions. Rather than bypassing this issue via information-lossy latent representations, we directly untangle the pixel-space trajectories by proposing Waypoint Diffusion Transformers (WiT). WiT factorizes the continuous vector field via intermediate semantic waypoints projected from pre-trained vision models. It effectively disentangles the generation trajectories by breaking the optimal transport into prior-to-waypoint and waypoint-to-pixel segments. Specifically, during the iterative denoising process, a lightweight generator dynamically infers these intermediate waypoints from the current noisy state. They then continuously condition the primary diffusion transformer via the Just-Pixel AdaLN mechanism, steering the evolution towards the next state, ultimately yielding the final RGB pixels. Evaluated on ImageNet 256x256, WiT beats strong pixel-space baselines, accelerating JiT training convergence by 2.2x. Code will be publicly released at https://github.com/hainuo-wang/WiT.git.