Learning to Solve Generative ODEs Beyond the Linear Span
About
Diffusion and flow generative models sample by integrating a learned ODE, but high quality still requires many sequential model evaluations. Solver learning reduces this cost by adapting scalar coefficients, timesteps, or both, while keeping the backbone model fixed. In this work, we identify a structural bottleneck in this update family: each step remains span-limited. Since the scalar-coefficient update lies in the span of buffered velocity evaluations, it can fit only the in-span component while leaving any out-of-span residual unreachable by scalar recombination alone. We propose SpanLift, a lightweight neural solver that augments scalar-coefficient updates with a spatial residual operator. SpanLift keeps a fixed base solver as an in-span prior and learns a spatial residual operator over the state and velocity buffer. The operator is trained by endpoint teacher matching, preserves the pretrained backbone, and adds no model NFEs. Empirically, the learned correction transfers across base solvers and is predominantly out-of-span. Across pixel-space diffusion, latent flow matching, and precipitation nowcasting, SpanLift achieves state-of-the-art few-step sampling. With only 3 NFE, it improves CIFAR-10 FID from 8.16 to 5.69 and ImageNet FID from 17.37 to 11.83.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Conditional Image Generation | ImageNet 64x64 | FID3.78 | 149 | |
| Unconditional Image Generation | CIFAR-10 32 x 32 | FID2.41 | 112 | |
| Unconditional Image Generation | FFHQ 64 x 64 | FID3.11 | 84 | |
| Image Generation | AFHQ 64x64 v2 (unconditional) | FID2.42 | 54 | |
| Unconditional Generation | AFHQ 64 x 64 v2 | FID2.42 | 48 | |
| Unconditional Image Generation | FFHQ unconditional 64x64 | FID3.11 | 45 | |
| Precipitation nowcasting | SEVIR (test) | CSI (16)63.12 | 23 | |
| Text-to-Image Generation | MJHQ (25K) | FID7.51 | 9 |