NVIDIA releases Gamma-World, a multi-agent world model supporting four-player collaboration and real-time 24 FPS

robot
Abstract generation in progress
The Crypto World News reports that NVIDIA has released a multi-agent generative world model called Gamma-World, which supports four-player collaboration and real-time action responses at 24 frames per second (fps). NVIDIA, in collaboration with researchers from Tsinghua University, the University of Toronto, and the Vector Institute, has broken through the long-standing limitations of virtual environment simulation being restricted to single or two-player interactions. The team has currently published the project page and paper, with plans to open-source the code and weights soon. The model introduces high-dimensional extensions of rotational position encoding and an information intermediary tagging mechanism, achieving for the first time the ability to extend from zero-shot two-player scenarios to four-player collaboration without retraining. To prevent the computational load from skyrocketing as the number of players increases, the solution incorporates a sparse central attention mechanism, successfully compressing the attention calculation cost between players to a linear scale. Evaluation results show that the new model significantly outperforms traditional slot-based and dense attention networks in video realism, controllability of action responses, and consistency among players.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • 8
  • 3
  • Share
Comment
Add a comment
Add a comment
BlackVelvetKey
· 10h ago
Four-person collaboration scenario, metaverse socializing has potential
View OriginalReply0
ReflectiveKey
· 10h ago
University of Toronto participation, Canada's AI strength has been underestimated
View OriginalReply0
GasfeeComplainer
· 10h ago
Joint release is better than fighting alone, benefiting both academia and industry.
View OriginalReply0
KiteRerouter
· 10h ago
NVIDIA is planning to replace all game NPCs with AI.
View OriginalReply0
DoNotTouchTheLiquidationLine.
· 10h ago
Linearizing costs is the key point; otherwise, the business game can't run properly.
View OriginalReply0
GateUser-a65ee044
· 10h ago
The information intermediary marker sounds like something from the communication protocol layer.
View OriginalReply0
RiskParachute
· 10h ago
Is the visual realism surpassing traditional networks? Is GAN about to be replaced?
View OriginalReply0
Lime-ColoredStop-LossLine
· 10h ago
Rotational position encoding? Just copied over the Transformer approach.
View OriginalReply0
  • Pinned