OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators
We propose OPSD-V, an on-policy self-distillation paradigm for post-training few-step autoregressive (AR) video diffusion models. Existing few-step AR video generators can produce long videos with low latency, but still suffer from error accumulation and weakened motion dynamics during long autoregressive rollout. OPSD-V reduces long-horizon degradation while preserving the original few-step inference path. The key idea is to introduce real long-video data as temporal context during training and use it to provide dense trajectory-level supervision. Specifically, the student follows the exact i
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- LinkedLinked via arxiv author · 85%Hongyu Liu →
“OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators”
- LinkedLinked via arxiv author · 85%Zhichun Wang →
“OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators”
- LinkedLinked via arxiv author · 85%Feng Gao →
“OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators”
- LinkedLinked via arxiv author · 85%Xuanhua He →
“OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators”
- LinkedLinked via arxiv author · 85%Yue Ma →
“OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators”
- LinkedLinked via arxiv author · 85%Ziyu Wan →
“OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators”
- LinkedLinked via arxiv author · 85%Yong Zhang →
“OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators”
- LinkedLinked via arxiv author · 85%Xiaoming Wei →
“OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators”