Artificer by Nua

Artificer, from Nua, is the company's first robotics decision model. It is designed to evaluate robot task progress and success, providing feedback that robot systems need during training, evaluation, and deployment. Assessing these outcomes from video often requires manual review or brittle, handcrafted objectives. Artificer is presented as frontier-level task evaluation in 4B parameters. Example cases include an arm that grasps the right object but places it in the wrong bin, and a robot that stops after completing only part of its task.