World in World: training-free control of frozen video world models for re-camera and revisits
udmrzn · x · 2026-09-13
The arXiv paper 'World in World' introduces a training-free inference-time interface for controlling frozen autoregressive video world models. It converts heterogeneous control evidence — source-video observations, target-view scene projections, geometry renderings, and retrieved generated states — into camera- and time-labelled visual-evidence K/V read through the model's native self-attention. A correspondence router pairs persistent point identities with geometry, while Evidence-wise attention CFG (EWA) regulates each auxiliary channel in a single denoising pass. Capabilities include camera-controlled rerendering, consistent long-horizon revisits, and human-motion tasks without retraining.
More from Multimodal
- NVIDIA open-sources NemotronLabs RE-USE, a 9.6M-param multilingual speech denoiser — NVIDIAAI · 2026-10-31
- AI-generated fake trailer imagines a 1999 Dino Crisis movie with retro VHS feel — noizlab_studio · 2026-10-31
- Spatial Hub aggregates ~5000 public COLMAP-compatible spatial datasets for world models — janusch_patas · 2026-10-31
- Single image to walkable 3D Gaussian splat scene using World Labs Atlas and WebGPU — XRarchitect · 2026-10-31
- AI video reimagines Suicide Squad's Joker as a rapper in a full music-video treatment — Silver_War5476 · 2026-10-31
- 60+ community-made ChatGPT slash-style words turn one word into a full visual style — heyshrutimishra · 2026-10-31