VIKI-Bench
收藏NIAID Data Ecosystem2026-05-02 收录
数据链接：
官方服务：
资源简介：
The VIKI Benchmark, designed to study visual reasoning in multi-agent embodied environments based on Maniskill. The benchmark is built upon three levels of task granularity: Agent Activation, Task Planning, and Trajectory Perception, including a diverse set of heterogeneous robots and multi-modal visual observations inputs from multiple viewpoints.
创建时间：
2025-05-13
