<a href=\"https://cdn-uploads.huggingface.co/production/uploads/68d63e44ccf464a96ac18bcb/i7zwXrV2WF6Hy56lbj3YW.png\" rel=\"nofollow\"><img src=\"https://cdn-uploads.huggingface.co/production/uploads/68d63e44ccf464a96ac18bcb/i7zwXrV2WF6Hy56lbj3YW.png\" alt=\"echo_forcing\"></a></p>\n","updatedAt":"2026-05-20T02:06:11.450Z","author":{"_id":"68d63e44ccf464a96ac18bcb","avatarUrl":"/avatars/01b93c353a7231fa9e053ba9edcb2b06.svg","fullname":"Weilun Feng","name":"wlfeng","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.45331040024757385},"editors":["wlfeng"],"editorAvatarUrls":["/avatars/01b93c353a7231fa9e053ba9edcb2b06.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.16003","authors":[{"_id":"6a0d16e665eb30f20d962ba0","name":"Mingqiang Wu","hidden":false},{"_id":"6a0d16e665eb30f20d962ba1","name":"Weilun Feng","hidden":false},{"_id":"6a0d16e665eb30f20d962ba2","name":"Zhefeng Zhang","hidden":false},{"_id":"6a0d16e665eb30f20d962ba3","name":"Haotong Qin","hidden":false},{"_id":"6a0d16e665eb30f20d962ba4","name":"Yuqi Li","hidden":false},{"_id":"6a0d16e665eb30f20d962ba5","name":"Guoxin Fan","hidden":false},{"_id":"6a0d16e665eb30f20d962ba6","name":"Xiaokun Liu","hidden":false},{"_id":"6a0d16e665eb30f20d962ba7","name":"Zhulin An","hidden":false},{"_id":"6a0d16e665eb30f20d962ba8","name":"Libo Huang","hidden":false},{"_id":"6a0d16e665eb30f20d962ba9","name":"Yongjun Xu","hidden":false},{"_id":"6a0d16e665eb30f20d962baa","name":"Chuanguang Yang","hidden":false}],"publishedAt":"2026-05-15T00:00:00.000Z","submittedOnDailyAt":"2026-05-20T00:00:00.000Z","title":"Echo-Forcing: A Scene Memory Framework for Interactive Long Video Generation","submittedOnDailyBy":{"_id":"68d63e44ccf464a96ac18bcb","avatarUrl":"/avatars/01b93c353a7231fa9e053ba9edcb2b06.svg","isPro":false,"fullname":"Weilun Feng","user":"wlfeng","type":"user","name":"wlfeng"},"summary":"Autoregressive video diffusion models enable open-ended generation through local attention and KV caching. However, existing training-free long-video optimization methods mainly focus on stable extension under a single prompt, making them difficult to handle interactive scenarios involving prompt switching, old scene forgetting, and historical scene recall. We identify the core bottleneck as the functional entanglement of historical KV states: stable anchors and recent dynamics are handled by the same cache policy, leading to outdated background contamination, delayed response to new prompts, and loss of long-range memory. To address this issue, we propose Echo-Forcing, a training-free scene memory framework specifically designed for interactive long video generation with three core mechanisms: (1) Hierarchical Temporal Memory, which decouples stable anchors, compressed history, and recent windows under relative RoPE; (2) Scene Recall Frames, which compresses historical scenes into spatially structured KV representations to support long-term recall; and (3) Difference-aware Memory Decay, which adaptively forgets conflicting tokens according to the discrepancy between old and new scenes. Based on these designs, Echo-Forcing uniformly supports smooth transitions, hard cuts, and long-range scene recall under a bounded cache budget. Extensive evaluations on VBench-Long further demonstrate that Echo-Forcing achieves the best overall performance in both long-video generation and interactive video generation settings. Our code is released in https://github.com/mingqiangWu/Echo-Forcing","upvotes":1,"discussionId":"6a0d16e665eb30f20d962bab","githubRepo":"https://github.com/mingqiangWu/Echo-Forcing","githubRepoAddedBy":"user","ai_summary":"Echo-Forcing addresses limitations in interactive long-video generation by decoupling historical memory and recent dynamics through hierarchical temporal memory, scene recall frames, and difference-aware memory decay mechanisms.","ai_keywords":["autoregressive video diffusion models","local attention","KV caching","training-free long-video optimization","functional entanglement","historical KV states","stable anchors","recent dynamics","relative RoPE","scene recall frames","difference-aware memory decay","long-range memory","VBench-Long"],"githubStars":15},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"68d63e44ccf464a96ac18bcb","avatarUrl":"/avatars/01b93c353a7231fa9e053ba9edcb2b06.svg","isPro":false,"fullname":"Weilun Feng","user":"wlfeng","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2605/2605.16003.md"}">
Echo-Forcing: A Scene Memory Framework for Interactive Long Video Generation
Authors: ,
,
,
,
,
,
,
,
,
,
Abstract
Echo-Forcing addresses limitations in interactive long-video generation by decoupling historical memory and recent dynamics through hierarchical temporal memory, scene recall frames, and difference-aware memory decay mechanisms.
AI-generated summary
Autoregressive video diffusion models enable open-ended generation through local attention and KV caching. However, existing training-free long-video optimization methods mainly focus on stable extension under a single prompt, making them difficult to handle interactive scenarios involving prompt switching, old scene forgetting, and historical scene recall. We identify the core bottleneck as the functional entanglement of historical KV states: stable anchors and recent dynamics are handled by the same cache policy, leading to outdated background contamination, delayed response to new prompts, and loss of long-range memory. To address this issue, we propose Echo-Forcing, a training-free scene memory framework specifically designed for interactive long video generation with three core mechanisms: (1) Hierarchical Temporal Memory, which decouples stable anchors, compressed history, and recent windows under relative RoPE; (2) Scene Recall Frames, which compresses historical scenes into spatially structured KV representations to support long-term recall; and (3) Difference-aware Memory Decay, which adaptively forgets conflicting tokens according to the discrepancy between old and new scenes. Based on these designs, Echo-Forcing uniformly supports smooth transitions, hard cuts, and long-range scene recall under a bounded cache budget. Extensive evaluations on VBench-Long further demonstrate that Echo-Forcing achieves the best overall performance in both long-video generation and interactive video generation settings. Our code is released in https://github.com/mingqiangWu/Echo-Forcing
Community
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2605.16003 in a model README.md to link it from this page.
Cite arxiv.org/abs/2605.16003 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2605.16003 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.