Rollout-Marginal Distillation for Long-Horizon Autoregressive Video Generation
Summary
This paper introduces Rollout-Marginal Distillation (RMD) for autoregressive video diffusion, where each generated chunk is scored independently against a chunk teacher to avoid error-compounding from imperfect temporal context, followed by video-level DMD to restore coherence. RMD reportedly maintains high visual quality far beyond its training horizon and outperforms video-level DMD baselines.
View Cached Full Text
Cached at: 10/05/26, 12:47 PM
Paper page - Rollout-Marginal Distillation for Long-Horizon Autoregressive Video Generation
Source: https://huggingface.co/papers/2609.37925 Published on Sep 29
·
Submitted byhttps://huggingface.co/cjeen
cjeenon Oct 5
Abstract
Autoregressive(AR)videodiffusionenableslow-latency,streamablevideogeneration,butpredictionerrorsoftenaccumulateoverlongrollouts.Trainingthegeneratoronitsownrolloutsexposesittotheseimperfecthistories.However,existingvideo-leveldistributionmatchingdistillation(DMD)scoresthewholerolloutjointly.Becauseachunkisevaluatedtogetherwithitspastandfuture,itscorrectioncanfavormatchingartifactsinthesurroundingcontextmerelytopreservetemporalconsistency.Toprovideaclearervisual-qualitysignal,weintroduceRollout-MarginalDistillation(RMD).RMDretainsthegeneratedhistoryforARpredictionbutscoreseachchunkindependentlyagainstachunkteacher,ensuringitsqualitycorrectionisnotcompromisedbyanimperfecttemporalcontext.Tocompensateforthelackoftemporalcontextinindependentchunkscoring,RMDsubsequentlyappliesvideo-levelDMDtorestoretemporalcoherence.ExtensiveexperimentsdemonstratethatRMDmaintainshighvisualqualityfarbeyonditstraininghorizonandoutperformsvideo-levelDMDbaselines.Codeandvideoresultsareavailableathttps://cjeen.github.io/RMD
View arXiv pageView PDFProject pageGitHubAdd to collection
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2609.37925 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2609.37925 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2609.37925 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
A^2RD: Agentic Autoregressive Diffusion for Long Video Consistency
A^2RD is a new paper introducing an Agentic Autoregressive Diffusion architecture for long video synthesis, achieving improved consistency and narrative coherence through a closed-loop self-improvement process.
AAD-1: Asymmetric Adversarial Distillation for One-Step Autoregressive Video Generation
AAD-1 introduces asymmetric adversarial distillation with phased training to achieve one-step autoregressive video generation, outperforming prior methods on VBench.
Causal-rCM: A Unified Teacher-Forcing and Self-Forcing Open Recipe for Autoregressive Diffusion Distillation in Streaming Video Generation and Interactive World Models
This paper introduces Causal-rCM, a unified teacher-forcing and self-forcing framework for autoregressive diffusion distillation in streaming video generation and interactive world models, achieving state-of-the-art performance with fast convergence.
OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generators
OPSD-V improves few-step autoregressive video diffusion models by using real long-video data as temporal context during training, providing dense trajectory-level supervision that enhances visual quality and motion dynamics without altering inference mechanisms.
Long Video Generation (4 minute read)
The article introduces A²RD, a novel architecture for generating consistent long videos using agentic autoregressive diffusion. It proposes a Retrieve–Synthesize–Refine–Update cycle and a new benchmark, LVBench-C, to address semantic drift in long-horizon video synthesis.