GeoState: View-Decoupled Geometry States Separate Scene Dynamics from Camera Motion in World Models
Tongji · hf · 2026-10-19
Researchers from Tongji University, HK PolyU and Tencent introduce GeoState, a view-decoupled geometric representation for predictive 3D world modeling.
- Existing world models use geometry foundation model (GFM) features as world states, but their view dependence entangles scene dynamics with camera motion, making state evolution hard to predict
- A scene tokenizer aggregates multi-view GFM features into a view-decoupled scene state, which a geometry decoder queries via camera rays
- An autoregressive world model predicts future scene states while camera motion is modeled separately, enabling consistent multi-view forecasting and cross-dataset generalization
More from Research
- MIST Framework Separates Execution from Memory to Diagnose Video World Model Failures — hkuhk · 2026-10-19
- Dusting the Phantom: Gaussian Allocation by Spatiotemporal Destination Recovers Fast-Moving Detail — Minhyeok Roh · 2026-10-19
- EvoHeir paper: fresh evaluation cuts harmful inherited updates in self-evolving LLM loops by 68–83% — MBZUAI · 2026-10-19
- ReaDy Uses Recurrent States as Dynamic Context for Speculative Decoding in Hybrid Linear Attention LLMs — Bohan Sun · 2026-10-19
- PrefEval: preference-following accuracy falls below 10% at just 10 turns for most LLMs — math_rachel · 2026-10-19
- ReasoningTrap: diagnosing reasoning rigidity with doctored variants of famous puzzles — math_rachel · 2026-10-19