PlayGen2: Real-Time Interactive World Model with Explicit 3D State
arXiv — cs.CV metadata · Jihai Zhang, Yuhan Zhou, Junyou Li, Deheng Ye, Qi Dou, Yu Cheng · Published on Readzer
Summary
PlayGen2 is a real-time interactive world model that separates learned 3D state evolution from neural rendering, using a persistent world memory with volumetric occupancy grids and identity-indexed 3D states. Tested on Doom, it learns geometry-aware player motion and player-NPC interactions, achieving ~160ms backend latency on a MacBook Pro M4 Pro. It supports zero-shot deployment on unseen maps, multiplayer via composition, and programmable gameplay, with code and weights publicly available.
This is a reviewed summary. Read the complete article at the original source.
Why it matters by industry
- Artificial intelligence
Relevant to AI models and embodied intelligence: introduces a world model with explicit 3D state for real-time interactive environments, advancing agent-environment interaction and transferability.