This framework achieves efficient few-step generation in diffusion models, suggesting a new approach to stable model training.
We propose SSD-CM, the first framework for consistency distillation on Mamba-2 (SSD) diffusion backbones. SSD-CM combines linear-complexity SSM layers with few-step generation, achieving compounding efficiency gains. We introduce SSD-Aware Consistency Training and Timestep-Adaptive State Initialization, two techniques that exploit Mamba-2's recurrent structure for stable distillation.
No takes yet. Share an insight, caveat, or question.
Hiroki Abe (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: