mlsanityData to 7 Oct 2026

Context Forcing: Consistent Autoregressive Video Generation with Long Context

arXiv 2602.06028 ↗First posted 5 Feb 2026cs.CVcited by 47

Not on the map. Why?

Abstract (arXiv)

Recent approaches to real-time long video generation typically employ streaming tuning strategies, attempting to train a long-context student using a short-context (memoryless) teacher. In these frameworks, the student performs long rollouts but receives supervision from a teacher limited to short 5-second windows. This structural discrepancy creates a critical student-teacher mismatch: the teacher's inability to access long-term history prevents it from guiding the student on global temporal dependencies, effectively capping the student's context length. To resolve this, we propose Context Forcing, a novel framework that trains a long-context student via a long-context teacher. By ensuring the teacher is aware of the full generation history, we eliminate the supervision mismatch, enabling the robust training of models capable of long-term consistency. To make this computationally feasible for extreme durations (e.g., 2 minutes), we introduce a context management system that transforms the linearly growing context into a Slow-Fast Memory architecture, significantly reducing visual redundancy. Extensive results demonstrate that our method enables effective context lengths exceeding 20 seconds -- 2 to 10 times longer than state-of-the-art methods like LongLive and Infinite-RoPE. By leveraging this extended context, Context Forcing preserves superior consistency across long durations, surpassing state-of-the-art baselines on various long video evaluation metrics.

Key facts. Cited by 47 papers in the map (data through 7 Oct 2026). Builds on 46 of its 49 references that are in the map. First posted on arXiv 5 Feb 2026.

Builds on

46 of 49 references.

References · 46

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  3. 2022Scalable Diffusion Models with Transformerscited by 5,931
  4. 2025Wan: Open and Advanced Large-Scale Video Generative Modelscited by 2,536
  5. 2024CogVideoX: Text-to-Video Diffusion Models with An Expert Transformercited by 2,198
  6. 2024HunyuanVideo: A Systematic Framework For Large Video Generative Modelscited by 1,597
  7. 2023ProlificDreamer: High-Fidelity and Diverse Text-to-3D Generation with Variational Score Distillationcited by 1,207
  8. 2023One-step Diffusion with Distribution Matching Distillationcited by 978
  9. 2024Improved Distribution Matching Distillation for Fast Image Synthesiscited by 648
  10. 2024Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusioncited by 634
  11. 2024Movie Gen: A Cast of Media Foundation Modelscited by 614
  12. 2025Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusioncited by 569
34 more
  1. 2024LTX-Video: Realtime Video Latent Diffusioncited by 549
  2. 2023InstaFlow: One Step is Enough for High-Quality Diffusion-Based Text-to-Image Generationcited by 387
  3. 2016Generating Videos with Scene Dynamicscited by 376
  4. 2024Pyramidal Flow Matching for Efficient Video Generative Modelingcited by 354
  5. 2025VBench-2.0: Advancing Video Generation Benchmark Suite for Intrinsic Faithfulnesscited by 289
  6. 2024Diffusion Models Are Real-Time Game Enginescited by 282
  7. 2024Fast High-Resolution Image Synthesis with Latent Adversarial Diffusion Distillationcited by 276
  8. 2025MAGI-1: Autoregressive Video Generation at Scalecited by 262
  9. 2025LongLive: Real-time Interactive Long Video Generationcited by 237
  10. 2025SkyReels-V2: Infinite-length Film Generative Modelcited by 218
  11. 2023LCM-LoRA: A Universal Stable-Diffusion Acceleration Modulecited by 211
  12. 2025Rolling Forcing: Autoregressive Long Video Diffusion in Real Timecited by 209
  13. 2024Autoregressive Video Generation without Vector Quantizationcited by 180
  14. 2025Self-Forcing++: Towards Minute-Scale High-Quality Video Generationcited by 176
  15. 2025Context as Memory: Scene-Consistent Interactive Long Video Generation with Memory Retrievalcited by 157
  16. 2025WorldMem: Long-term Consistent World Simulation with Memorycited by 152
  17. 2025WorldPlay: Towards Long-Term Geometric Consistency for Real-Time Interactive World Modelingcited by 152
  18. 2024FIFO-Diffusion: Generating Infinite Videos from Text without Trainingcited by 134
  19. 2024VidProM: A Million-scale Real Prompt-Gallery Dataset for Text-to-Video Diffusion Modelscited by 124
  20. 2025Long-Context Autoregressive Video Modeling with Next-Frame Predictioncited by 119
  21. 2016Video Pixel Networkscited by 108
  22. 2025RELIC: Interactive Video World Model with Long-Horizon Memorycited by 89
  23. 2024FreeLong: Training-Free Long Video Generation with SpectralBlend Temporal Attentioncited by 87
  24. 2025Autoregressive Adversarial Post-Training for Real-Time Interactive Video Generationcited by 84
  25. 2024Phased Consistency Modelscited by 83
  26. 2025Infinity-RoPE: Action-Controllable Infinite Video Generation Emerges From Autoregressive Self-Rolloutcited by 78
  27. 2025MotionStream: Real-Time Video Generation with Interactive Motion Controlscited by 71
  28. 2025Stable Video Infinity: Infinite-Length Video Generation with Error Recyclingcited by 68
  29. 2025Sekai: A Video Dataset towards World Explorationcited by 64
  30. 2025RIFLEx: A Free Lunch for Length Extrapolation in Video Diffusion Transformerscited by 63
  31. 2025UltraVideo: High-Quality UHD Video Dataset with Comprehensive Captionscited by 52
  32. 2025StreamDiT: Real-Time Streaming Text-to-Video Generationcited by 43
  33. 2025FreeLong++: Training-Free Long Video Generation via Multi-band SpectralFusioncited by 15
  34. 2025TinyHistory: Lightweight Video History Embeddings via Two-Stage Context Learningcited by 13

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 47.

47 papers.

Sort:

By area:

  1. 2026LongLive-2.0: An NVFP4 Parallel Infrastructure for Long Video Generationcited by 25
  2. 2026MemRoPE: Training-Free Infinite Video Generation via Evolving Memory Tokenscited by 20
  3. 2026Relax Forcing: Relaxed KV-Memory for Consistent Long Video Generationcited by 20
  4. 2026Self Gradient Forcing: Native Long Video Extrapolationcited by 16
  5. 2026ABot-World-0: Infinite Interactive World Rollout on a Single Desktop GPUcited by 13
  6. 2026Mode Seeking meets Mean Seeking for Fast Long Video Generationcited by 12
  7. 2026LongLive-RAG: A General Retrieval-Augmented Framework for Long Video Generationcited by 12
  8. 2026Addressable Memory for Video World Modelscited by 11
  9. 2026SVG-EAR: Parameter-Free Linear Compensation for Sparse Video Generation via Error-aware Routingcited by 9
  10. 2026AvatarForcing: One-Step Streaming Talking Avatars via Local-Future Sliding-Window Denoisingcited by 6
  11. 2026A2RD: Agentic Autoregressive Diffusion for Long Video Consistencycited by 6
  12. 2026RAVEN: Real-time Autoregressive Video Extrapolation with Consistency-model GRPOcited by 6
35 more
  1. 2026Advancing Narrative Long Video Generation via Training-Free Identity-Aware Memorycited by 6
  2. 2026Towards Interactive Video World Modeling: Frontiers, Challenges, Benchmarks, and Future Trendscited by 6
  3. 2026Echo-Infinity: Learning Evolving Memory for Real-Time Infinite Video Generationcited by 6
  4. 2026Compression and Retrieval: Implicit Memory Retrieval for Video World Modelscited by 6
  5. 2026Grounded Forcing: Bridging Time-Independent Semantics and Proximal Dynamics in Autoregressive Video Synthesiscited by 5
  6. 2026Head Forcing: Long Autoregressive Video Generation via Head Heterogeneitycited by 5
  7. 2026DIM-WAM: World-Action Modeling with Diverse Historical Event Memorycited by 5
  8. 2026Context-Matched Distillation: Teacher Causality for Autoregressive Video Distillationcited by 5
  9. 2026AnyRecon: Arbitrary-View 3D Reconstruction with Video Diffusion Modelcited by 4
  10. 2026AlayaWorld: Long-Horizon and Playable Video World Generationcited by 4
  11. 2026DistillAlign: Coordinating Mode Covering and Mode Seeking in Autoregressive Video Distillationcited by 4
  12. 2026EverAnimate: Minute-Scale Human Animation via Latent Flow Restorationcited by 3
  13. 2026Goodbye Drift: Anchored Tree Sampling for Long-Horizon Video-to-Video Generationcited by 2
  14. 2026LiveBand: Live Accompaniment Generation in the Audio Domaincited by 2
  15. 2026StreamHOI: Interaction-aware Temporal Memory Adaptation for Streaming HOI Video Generationcited by 2
  16. 2026RECAP-Forcing: Retaining Content Appearances for Long Video Generationcited by 2
  17. 2026LayerRecall: A State-Conditioned Memory Router for Long-Horizon Consistency in Video Generationcited by 2
  18. 2026Delta Forcing: Trust Region Steering for Interactive Autoregressive Video Generationcited by 1
  19. 2026Focused Forcing: Content-Aware Per-Frame KV Selection for Efficient Autoregressive Video Diffusioncited by 1
  20. 2026Quantized Keys Steal Attention: Bias Correction for KV-Cache Compression in Video Diffusioncited by 1
  21. 2026AnchorEdit: Maintaining Temporal Consistency in Multi-turn Image Editing via Causal Memorycited by 1
  22. 2026Cycle-World: Mitigating Error Accumulation in Long-term Video World Models via Reverse-Prediction Cycle Consistencycited by 1
  23. 2026LiveAnimate: Stable Long-Form Streaming Human Animation in Real-Timecited by 1
  24. 2026Do Video Generators Track the World Across Segments? A Benchmark and Method for World-State Reasoning in Video Continuationcited by 1
  25. 2026Precise Editing and Flexible Referencing for Interactable Worldscited by 1
  26. 2026EchoTorrent: Towards Swift, Sustained, and Streaming Multi-Modal Video Generationcited by 0
  27. 2026ReCoSplat: Online Feed-Forward Gaussian Splatting via Render-and-Comparecited by 0
  28. 2026AtlasVid: Efficient Ultra-High-Resolution Long Video Generation via Decoupled Global-Local Modelingcited by 0
  29. 2026EditStream: A Unified Autoregressive Framework for Interactive Video Generation and Editingcited by 0
  30. 2026ViRDM: Taming Representation Distribution Matching for Few-Step Causal Video Generationcited by 0
  31. 2026From Scores to Samples: Elastic Forcing for Autoregressive Video Generationcited by 0
  32. 2026LongTake: Learning to Sustain Dynamics in Long-Horizon Video Generationcited by 0
  33. 2026Learning to Retrieve: Internalizing Memory Retrieval for Video World Modelscited by 0
  34. 2026Conditional Residual Prediction: Improving Autoregressive Video Diffusion without a Bidirectional Teachercited by 0
  35. 2026Connected Self Forcing: Beyond Local Learning in Video Autoregressioncited by 0