Open project lets users test AI models on simulated robot tasks

A user shares an open project for connecting a model to a simulated robot and measuring how well it completes a task. In the described loop, SenseNova-RoboRSI-27B-Preview receives a task and camera images, chooses a movement, then uses feedback to decide what to do next. The post cites a 26.13 average score and 20.83% task success on RoboDojo, plus 96.1% success with direct control on LIBERO-Pro.