Research focus
Example task
Current state
pump 60% · outlet 18.4°C
Pump adjustment
60% → 70%
Predicted response
outlet 17.7°C
Observed response
outlet 17.9°C
Synthetic example values. The same pump, pipes, and heat exchanger appear in each stage. Dashed marks indicate prediction; solid marks indicate observation.
Illustrative task record. It does not represent a published evaluation result.
What the task measures.
The record includes the current state, the requested pump setting, the model’s predicted response, and the observed system response. Predictions are shown with dashed lines; observations use solid lines.
Keeping the same physical objects in each frame makes changes in the system state easier to compare. This focus includes world models, manufacturing, robotics, and industrial control.
Discuss physical systems evaluation.
Evaluation engagements are custom. We agree on the task, inputs, outcomes, and scoring before a run begins.