mlsanityData to 14 Oct 2026

Rolling Forcing: Autoregressive Long Video Diffusion in Real Time

arXiv 2509.25161 ↗First posted 29 Sep 2025cs.CVcited by 234

On the map: World models → Video-generation world models and simulators

also in 1 more
Abstract (arXiv)

Streaming video generation, as one fundamental component in interactive world models and neural game engines, aims to generate high-quality, low-latency, and temporally coherent long video streams. However, most existing work suffers from severe error accumulation that often significantly degrades the generated stream videos over long horizons. We design Rolling Forcing, a novel video generation technique that enables streaming long videos with minimal error accumulation. Rolling Forcing comes with three novel designs. First, instead of iteratively sampling individual frames, which accelerates error propagation, we design a joint denoising scheme that simultaneously denoises multiple frames with progressively increasing noise levels. This design relaxes the strict causality across adjacent frames, effectively suppressing error growth. Second, we introduce the attention sink mechanism into the long-horizon stream video generation task, which allows the model to keep key value states of initial frames as a global context anchor and thereby enhances long-term global consistency. Third, we design an efficient training algorithm that enables few-step distillation over largely extended denoising windows. This algorithm operates on non-overlapping windows and mitigates exposure bias conditioned on self-generated histories. Extensive experiments show that Rolling Forcing enables real-time streaming generation of multi-minute videos on a single GPU, with substantially reduced error accumulation.

Key facts. Cited by 234 papers in the map (data through 14 Oct 2026). Placed in Video-generation world models and simulators, one of 11 directions in World models. Builds on 54 of its 54 references that are in the map. First posted on arXiv 29 Sep 2025.

Builds on

54 of 54 references.

Closest ancestors · 9

Same direction, newest first

  1. 2025MAGI-1: Autoregressive Video Generation at Scale19 May
  2. 2025Wan: Open and Advanced Large-Scale Video Generative Models26 Mar
  3. 2024Diffusion Models Are Real-Time Game Engines27 Aug
  4. 2024StreamingT2V: Consistent, Dynamic, and Extendable Long Video Generation from Text21 Mar
  5. 2024Genie: Generative Interactive Environments23 Feb
  6. 2023Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets25 Nov
  7. 2022Scalable Diffusion Models with Transformers19 Dec
  8. 2022Flow Matching for Generative Modeling6 Oct
  9. 2021VideoGPT: Video Generation using VQ-VAE and Transformers20 Apr

Other references · 45

  1. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,703
  2. 2022Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flowcited by 3,700
  3. 2023Efficient Streaming Language Models with Attention Sinkscited by 2,186
  4. 2023VBench: Comprehensive Benchmark Suite for Video Generative Modelscited by 1,822
  5. 2022Imagen Video: High Definition Video Generation with Diffusion Modelscited by 1,778
  6. 2022Make-A-Video: Text-to-Video Generation without Text-Video Datacited by 1,742
  7. 2024HunyuanVideo: A Systematic Framework For Large Video Generative Modelscited by 1,660
  8. 2023Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Modelscited by 1,476
  9. 2023One-step Diffusion with Distribution Matching Distillationcited by 1,019
  10. 2022CogVideo: Large-scale Pretraining for Text-to-Video Generation via Transformerscited by 985
  11. 2024Improved Distribution Matching Distillation for Fast Image Synthesiscited by 684
  12. 2024Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusioncited by 681
33 more
  1. 2024Movie Gen: A Cast of Media Foundation Modelscited by 644
  2. 2025Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusioncited by 625
  3. 2023VideoPoet: A Large Language Model for Zero-Shot Video Generationcited by 492
  4. 2024From Slow Bidirectional to Fast Autoregressive Video Diffusion Modelscited by 433
  5. 2024Pyramidal Flow Matching for Efficient Video Generative Modelingcited by 366
  6. 2023Photorealistic Video Generation with Diffusion Modelscited by 306
  7. 2025SkyReels-V2: Infinite-length Film Generative Modelcited by 229
  8. 2019Scaling Autoregressive Video Modelscited by 190
  9. 2025Diffusion Adversarial Post-Training for One-Step Video Generationcited by 151
  10. 2025Frame Context Packing and Drift Prevention in Next-Frame-Prediction Video Diffusion Modelscited by 143
  11. 2024FIFO-Diffusion: Generating Infinite Videos from Text without Trainingcited by 137
  12. 2024VidProM: A Million-scale Real Prompt-Gallery Dataset for Text-to-Video Diffusion Modelscited by 131
  13. 2025Test-Time Training Done Rightcited by 130
  14. 2019Generalization in Generation: A closer look at Exposure Biascited by 123
  15. 2025Long-Context Autoregressive Video Modeling with Next-Frame Predictioncited by 123
  16. 2025Long Context Tuning for Video Generationcited by 113
  17. 2024Rolling Diffusion Modelscited by 104
  18. 2024Loong: Generating Minute-level Long Videos with Autoregressive Language Modelscited by 99
  19. 2025Autoregressive Adversarial Post-Training for Real-Time Interactive Video Generationcited by 90
  20. 2024MovieDreamer: Hierarchical Generation for Coherent Long Visual Sequencecited by 89
  21. 2025Seaweed-7B: Cost-Effective Training of Video Generation Foundation Modelcited by 82
  22. 2023ARTboldsymbol·V: Auto-Regressive Text-to-Video Generation with Diffusion Modelscited by 76
  23. 2024Progressive Autoregressive Video Diffusion Modelscited by 68
  24. 2024VideoStudio: Generating Consistent-Content and Multi-Scene Videoscited by 64
  25. 2025AR-Diffusion: Asynchronous Video Generation with Auto-Regressive Diffusioncited by 55
  26. 2024MarDini: Masked Autoregressive Diffusion for Video Generation at Scalecited by 49
  27. 2025StreamDiT: Real-Time Streaming Text-to-Video Generationcited by 48
  28. 2024ARLON: Boosting Diffusion Transformers with Autoregressive Models for Long Video Generationcited by 41
  29. 2024DreamFactory: Pioneering Multi-Scene Long Video Generation with a Multi-Agent Frameworkcited by 35
  30. 2024StoryAgent: Customized Storytelling Video Generation via Multi-Agent Collaborationcited by 31
  31. 2024TALC: Time-Aligned Captions for Multi-Scene Text-to-Video Generationcited by 25
  32. 2025Macro-from-Micro Planning for High-Quality and Parallelized Autoregressive Long Video Generationcited by 21
  33. 2024Synchronized Video Storytelling: Generating Video Narrations with Structured Storylinecited by 19

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 24; 2026: 210.

234 papers.

Sort:

By area:

  1. 2025Memory in the Age of AI Agentscited by 303
  2. 2025Self-Forcing++: Towards Minute-Scale High-Quality Video Generationcited by 192
  3. 2026Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Interactive Video Generationcited by 184
  4. 2025WorldPlay: Towards Long-Term Geometric Consistency for Real-Time Interactive World Modelingcited by 170
  5. 2026DreamDojo: A Generalist Robot World Model from Large-Scale Human Videoscited by 135
  6. 2025RELIC: Interactive Video World Model with Long-Horizon Memorycited by 96
  7. 2025Infinity-RoPE: Action-Controllable Infinite Video Generation Emerges From Autoregressive Self-Rolloutcited by 86
  8. 2025Reward Forcing: Efficient Streaming Video Generation with Rewarded Distribution Matching Distillationcited by 86
  9. 2025Deep Forcing: Training-Free Long Video Generation with Deep Sink and Participative Compressioncited by 85
  10. 2026Helios: Real Real-Time Long Video Generation Modelcited by 65
  11. 2025Hunyuan-GameCraft-2: Instruction-following Interactive Game World Modelcited by 61
  12. 2026Context Forcing: Consistent Autoregressive Video Generation with Long Contextcited by 59
38 more
  1. 2026Causal Forcing++: Scalable Few-Step Autoregressive Diffusion Distillation for Real-Time Interactive Video Generationcited by 56
  2. 2025VideoSSM: Autoregressive Long Video Generation with Hybrid State-Space Memorycited by 42
  3. 2026LoL: Longer than Longer, Scaling Video Generation to Hourcited by 32
  4. 2026Rolling Sink: Bridging Limited-Horizon Training and Open-Ended Testing in Autoregressive Video Diffusioncited by 32
  5. 2026LongLive-2.0: An NVFP4 Parallel Infrastructure for Long Video Generationcited by 30
  6. 2026ABot-M0.5: Unified Mobility-and-Manipulation World Action Modelcited by 30
  7. 2026LiveWorld: Simulating Out-of-Sight Dynamics in Generative Video World Modelscited by 28
  8. 2026ShotStream: Streaming Multi-Shot Video Generation for Interactive Storytellingcited by 28
  9. 2026MemRoPE: Training-Free Infinite Video Generation via Evolving Memory Tokenscited by 27
  10. 2026MosaicMem: Hybrid Spatial Memory for Controllable Video World Modelscited by 27
  11. 2026VideoMemory: Toward Consistent Video Generation via Memory Integrationcited by 25
  12. 2026Relax Forcing: Relaxed KV-Memory for Consistent Long Video Generationcited by 24
  13. 2026RealWonder: Real-Time Physical Action-Conditioned Video Generationcited by 22
  14. 2025Memorize-and-Generate: Towards Long-Term Consistency in Real-Time Video Generationcited by 20
  15. 2026Efficient Autoregressive Video Diffusion with Dummy Headcited by 20
  16. 2026Fast Autoregressive Video Diffusion and World Models with Temporal Cache Compression and Sparse Attentioncited by 18
  17. 2026Next Forcing: Causal World Modeling with Multi-Chunk Predictioncited by 18
  18. 2026PackForcing: Short Video Training Suffices for Long Video Sampling and Long Context Inferencecited by 17
  19. 2026Self Gradient Forcing: Native Long Video Extrapolationcited by 17
  20. 2025FloodDiffusion: Tailored Diffusion Forcing for Streaming Motion Generationcited by 16
  21. 2025BAgger: Backwards Aggregation for Mitigating Drift in Autoregressive Video Diffusion Modelscited by 16
  22. 2026Past- and Future-Informed KV Cache Policy with Salience Estimation in Autoregressive Video Diffusioncited by 16
  23. 2026MultiGen: Level-Design for Editable Multiplayer Worlds in Diffusion Game Enginescited by 16
  24. 2026OmniForcing: Unleashing Real-time Joint Audio-Visual Generationcited by 16
  25. 2026LongLive-RAG: A General Retrieval-Augmented Framework for Long Video Generationcited by 16
  26. 2026ABot-World-0: Infinite Interactive World Rollout on a Single Desktop GPUcited by 16
  27. 2026Pathwise Test-Time Correction for Autoregressive Long Video Generationcited by 15
  28. 2026AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memoriescited by 15
  29. 2026HiAR: Efficient Autoregressive Long Video Generation via Hierarchical Denoisingcited by 15
  30. 2026Forcing-KV: Hybrid KV Cache Compression for Efficient Autoregressive Video Diffusion Modelscited by 15
  31. 2026WorldKV: Efficient World Memory with World Retrieval and Compressioncited by 15
  32. 2026One-Forcing: Towards Stable One-Step Autoregressive Video Generationcited by 15
  33. 2026LiveEdit: Towards Real-Time Diffusion-Based Streaming Video Editingcited by 15
  34. 2026Wonder: Video World Model Done Bettercited by 15
  35. 2026Addressable Memory for Video World Modelscited by 15
  36. 2026OPSD-V: On-Policy Self-Distillation for Post-Training Few-Step Autoregressive Video Generatorscited by 14
  37. 2026Alaya-EVOKE: From Linear-Scaling Supervision to Endless Worldcited by 14
  38. 2026Code World Model: Coding Agent as World Braincited by 14

50 most cited of 234 All 234 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2026Equilibrium Forcing: Adaptive Video Generation Without Noise Conditioning
  2. 2026Streaming Autoregressive Video Generation via Diagonal Distillation
  3. 2026Causal-rCM: A Unified Teacher-Forcing and Self-Forcing Open Recipe for Autoregressive Diffusion Distillation in Streaming Video Generation and Interactive World Models
  4. 2025Playing with Transformer at 30+ FPS via Next-Frame Diffusion
  5. 2025MotionStream: Real-Time Video Generation with Interactive Motion Controls
  6. 2026S2PD: Serial-to-Parallel Diffusion for Physically and Logically Consistent Video Generation
  7. 2025Tuning-Free Long Video Generation via Global-Local Collaborative Diffusion
  8. 2025LongLive: Real-time Interactive Long Video Generation