option
Home
Flash News
Content
RalphEvans
RalphEvans
June 2, 2026

NVIDIA with Tsinghua, U. Toronto, and Vector Institute launched Gamma World a multi-agent world model. It uses Simplex Rotary Agent Encoding for equal agent representation and Sparse Hub Attention to achieve linear complexity enabling 24FPS real time simulation. In multi player Minecraft it reduced FVD by over 40% and was successfully applied to real dual arm robot collaboration tasks.

NVIDIA with Tsinghua, U. Toronto, and Vector Institute launched Gamma World a multi-agent world model. It uses Simplex Rotary Agent Encoding for equal agent representation and Sparse Hub Attention to achieve linear complexity enabling 24FPS real time simulation. In multi player Minecraft it reduced FVD by over 40% and was successfully applied to real dual arm robot collaboration tasks.
Comments (0)
0/300
OR