DART: Distillation-Aware Reparameterization for Training-Free LoRA Reuse in Few-Step Video Diffusion Models

RESUMO

Step distillation reduces the cost of video generation, but reusing a LoRA trained for longer trajectory can alter its functional effect or degrade target quality. Static parameter compatibility offers one perspective on this problem; our observations show that similar measured geometry coexist with different adapter behavior under a shortened denoising schedule. We propose DART, a training-free method that combines low-rank coordinate transport with target-schedule response calibration using forward evaluations and no source training videos. On a four-step Wan2.2 target, DART-F improves the joint quality score from 0.9029 to 0.9227 and changes macro functional retention from -0.4644 to +0.1349. Component analysis shows that calibration accounts for most of the improvement, while coordinate transport provides complementary gains when combined with calibration. Adapter-level results reveal positive functional effects for some adapters with strong attenuation and reduced negative effects for others. Evaluations on two additional targets show the same aggregate trend. These results motivate evaluating distilled-model LoRA reuse jointly through functional preservation and negative-transfer avoidance, without assuming recovery for every adapter.