DySink: Dynamic Frame Sinks for Autoregressive Long Video Generation

• URL: http://arxiv.org/abs/2605.21028v1
• Date: Wed, 20 May 2026 11:01:01 GMT
• Title: DySink: Dynamic Frame Sinks for Autoregressive Long Video Generation
• Abstract summary: We propose DySink, a retrieval-based framework that maintains a compact memory bank and selects visually relevant historical frames as dynamic frame sinks.<n> Experiments on minute-long videos show that DySink consistently improves dynamic degree over strong baselines while also achieving higher temporal quality.
• Score: 51.58559206569209
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Autoregressive long video generation often adopts bounded-memory streaming for efficiency, typically combining local windows for short-term continuity with static early-frame sinks as long-range anchors. However, this fixed allocation keeps early frames cached even when the current visual state has substantially diverged from them, while discarding potentially more relevant intermediate history. As a result, the retained long-range context may become less adaptive and bias generation toward outdated cues; in severe cases, RoPE-induced phase re-alignment can homogenize inter-head attention and cause sink collapse, where content regresses toward sink frames. We propose DySink, a retrieval-based framework that maintains a compact memory bank and selects visually relevant historical frames as dynamic frame sinks. DySink couples adaptive retrieval with a sink anomaly gate, which detects excessive inter-head consensus over retrieved context and suppresses collapse-prone context. Experiments on minute-long videos show that DySink consistently improves dynamic degree over strong baselines while also achieving higher temporal quality. The code and model weights will be released at https://github.com/yebo0216best/DySink.

Related papers

• TetherCache: Stabilizing Autoregressive Long-Form Video Generation with Gated Recall and Trusted Alignment [51.33418612284208]
  We propose TetherCache, a training-free and plug-and-play cache management strategy for drift-resistant long video generation.<n>Gated Recall with Attention-Diversity Balancing selects long-range memory frames using a gated score.<n>TAME lightly edits newly recalled memory tokens by aligning their statistics to a trusted context distribution.
  arXiv  Detail & Related papers  (2026-06-11T08:16:08Z)
• FadeMem: Distance-Aware Memory Consolidation for Autoregressive Video Diffusion [59.207505503284715]
  FadeMem organizes historical KV blocks into a temporal hierarchy under a fixed cache budget.<n>New history is inserted as fine-grained entries, while older adjacent entries are progressively merged.<n>Experiments show improved subject consistency, background stability, and temporal coherence over existing bounded-cache strategies.
  arXiv  Detail & Related papers  (2026-06-09T10:22:18Z)
• Steady-Forcing: Balancing Spatial Persistence and Motion Continuity in Long-Horizon Nature Video Diffusion [1.1666234644810893]
  Autoregressive video diffusion models enable streaming generation but often degrade over long rollouts.<n>We study this stability-motion trade-off in fixed-camera long-horizon nature video generation.<n>We propose Steady-Forcing, a memory and training framework combining a persistent visual anchor (V-Sink), an exponential moving-average motion memory (EMA-Sink), block-relative temporal encoding, periodic cache purification, and distillation from a Wan2.1-14B teacher.
  arXiv  Detail & Related papers  (2026-06-02T07:11:50Z)
• Attend Locally, Remember Linearly: Linear Attention as Cross-Frame Memory for Autoregressive Video Diffusion [61.57938553036056]
  ARL2 is a hybrid attention module that replaces quadratic cross-frame attention with a fixed-size recurrent state.<n>We decompose self-attention into two branches: an intra-frame softmax branch for spatial detail and local dependencies, and an inter-frame recurrent linear branch that maintains a fixed-size state for streaming context.<n>With 75% of layers replaced by hybrid linear attention, the model achieves up to 2.26 wall-clock speedup and 54% memory reduction, while maintaining comparable quality with improving temporal consistency.
  arXiv  Detail & Related papers  (2026-05-15T19:33:45Z)
• Echo-Forcing: A Scene Memory Framework for Interactive Long Video Generation [48.476317015122625]
  Echo-Forcing is a training-free scene memory framework for interactive long video generation.<n>It supports smooth transitions, hard cuts, and long-range scene recall under a bounded cache budget.
  arXiv  Detail & Related papers  (2026-05-15T14:33:09Z)
• SWIFT: Prompt-Adaptive Memory for Efficient Interactive Long Video Generation [31.105007908298003]
  Streaming long-video generation faces a central challenge in continuous semantic switching.<n>Current approaches rely on cache rebuilding at prompt boundaries or fixed memory budgets.<n>We present SWIFT, a training-free framework for multi-prompt long-video generation.
  arXiv  Detail & Related papers  (2026-05-10T09:37:56Z)
• Relax Forcing: Relaxed KV-Memory for Consistent Long Video Generation [73.84423888025171]
  Autoregressive (AR) video diffusion has recently emerged as a promising paradigm for long video generation.<n>We show that extending generation to minute-scale horizons remains challenging due to progressive temporal degradation.<n>We introduce Relax Forcing, a structured temporal memory mechanism for AR diffusion.
  arXiv  Detail & Related papers  (2026-03-22T18:59:24Z)
• S2GS: Streaming Semantic Gaussian Splatting for Online Scene Understanding and Reconstruction [57.07346645250984]
  Streaming Semantic Gaussian Splatting (S2GS) is a strictly causal, incremental 3D Gaussian semantic field framework.<n>It does not leverage future frames and continuously updates scene geometry, appearance, and instance-level semantics without reprocessing historical frames.<n>S2GS matches or outperforms strong offline baselines on joint reconstruction-and-understanding benchmarks.
  arXiv  Detail & Related papers  (2026-03-15T05:48:55Z)
• MemRoPE: Training-Free Infinite Video Generation via Evolving Memory Tokens [34.228121359393775]
  We introduce MemRoPE, a training-free framework with two co-designed components.<n>Online RoPE Indexing caches unrotated keys and applies positional embeddings dynamically at attention time.<n>MemRoPE outperforms existing methods in temporal coherence, visual fidelity, and subject consistency across minute- to hour-scale generation.
  arXiv  Detail & Related papers  (2026-03-12T23:14:16Z)
• LoL: Longer than Longer, Scaling Video Generation to Hour [50.945885467651216]
  This work achieves the first demonstration of real-time, streaming, and infinite-length video generation with little quality decay.<n>As an illustration, we generate continuous videos up to 12 hours in length, which, to our knowledge, is among the longest publicly demonstrated results in streaming video generation.
  arXiv  Detail & Related papers  (2026-01-23T17:21:35Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.