{"paper":{"title":"WorldPlay: Towards Long-Term Geometric Consistency for Real-Time Interactive World Modeling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"WorldPlay generates long-horizon 720p video at 24 FPS while preserving geometric consistency through rebuilt context memory in a streaming diffusion model.","cross_cats":["cs.GR"],"primary_cat":"cs.CV","authors_text":"Chunchao Guo, Haiyu Zhang, Haoyuan Wang, Junta Wu, Jun Zhang, Tengfei Wang, Wenqiang Sun, Yunhong Wang, Zehan Wang, Zhenwei Wang","submitted_at":"2025-12-16T17:22:46Z","abstract_excerpt":"This paper presents WorldPlay, a streaming video diffusion model that enables real-time, interactive world modeling with long-term geometric consistency, resolving the trade-off between speed and memory that limits current methods. WorldPlay draws power from three key innovations. 1) We use a Dual Action Representation to enable robust action control in response to the user's keyboard and mouse inputs. 2) To enforce long-term consistency, our Reconstituted Context Memory dynamically rebuilds context from past frames and uses temporal reframing to keep geometrically important but long-past fram"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"WorldPlay generates long-horizon streaming 720p video at 24 FPS with superior consistency, comparing favorably with existing techniques and showing strong generalization across diverse scenes.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The Reconstituted Context Memory and Context Forcing methods actually preserve long-range geometric information and prevent error drift in practice without introducing new artifacts or requiring impractical memory.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"WorldPlay uses dual action representation, reconstituted context memory, and context forcing distillation to produce consistent 720p streaming video at 24 FPS for interactive world modeling.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"WorldPlay generates long-horizon 720p video at 24 FPS while preserving geometric consistency through rebuilt context memory in a streaming diffusion model.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"fc4db54fe6af1dfc4cd1ab8e55b4ff3e745b14a4fe84bcb859d372b29d366ad1"},"source":{"id":"2512.14614","kind":"arxiv","version":1},"verdict":{"id":"4c14e863-3cab-4354-b7d5-ec17a064a9ee","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T14:25:17.859701Z","strongest_claim":"WorldPlay generates long-horizon streaming 720p video at 24 FPS with superior consistency, comparing favorably with existing techniques and showing strong generalization across diverse scenes.","one_line_summary":"WorldPlay uses dual action representation, reconstituted context memory, and context forcing distillation to produce consistent 720p streaming video at 24 FPS for interactive world modeling.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The Reconstituted Context Memory and Context Forcing methods actually preserve long-range geometric information and prevent error drift in practice without introducing new artifacts or requiring impractical memory.","pith_extraction_headline":"WorldPlay generates long-horizon 720p video at 24 FPS while preserving geometric consistency through rebuilt context memory in a streaming diffusion model."},"references":{"count":81,"sample":[{"doi":"","year":2024,"title":"Diffusion for world modeling: Visual details matter in atari","work_id":"d9676607-838e-4c01-a91d-fdd83d1c181f","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Ac3d: Analyzing and improving 3d camera control in video diffusion trans- formers","work_id":"04724e59-193c-4821-ae21-21ebd6c9e842","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Navigation world models","work_id":"13a8b5e7-df85-4cad-9645-8e684faae06e","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Uni3c: Unifying precisely 3d-enhanced camera and hu- man motion controls for video generation.arXiv preprint arXiv:2504.14899, 2025","work_id":"c4cc0875-d298-400d-a5cd-f9de3bb813d1","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Gamegen-x: Interactive open-world game video generation","work_id":"5f3b923c-3fdc-468a-8951-94c4ff7b1e29","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":81,"snapshot_sha256":"da74a3586c3b9bd6d657fe9bfe6721d8e959bb6a50d90f19a7eac5676244b3b0","internal_anchors":17},"formal_canon":{"evidence_count":3,"snapshot_sha256":"d6ec8d5fc53940cdf3d0b7f087d734775aaed757aa639cb278e82ad6066e6ebf"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}