Case Study

Open Drawers

Case 1: Both systems succeed.

Case 2: Baseline system fails due to detection noise.

Case 3: Baseline system fails due to lack of geometric reasoning.

Case 4: Our system fails due to perception API errors.

Fetch Tools

Case 1: Both systems succeed.

Case 2: Baseline system fails due to detection noise.

Case 3: Baseline system fails due to lack of geometric reasoning.

Case 4: Our system fails due to detection noise.