Affordance Grounding
A task in embodied AI that requires a model to identify the possible actions an environmental object affords and output the corresponding image region, such as a bounding box or segmentation mask. It bridges visual perception and physical manipulation.
Timeline (last 90 days)
BEAKER's task types include affordance localization, trajectory planning, equipment/state/safety recognition, apparatus setup, operation and step reasoning, and temporal procedure reasoning.
Future model improvements for laboratories should prioritize filling in "action grounding" capabilities such as affordance localization and trajectory planning, rather than continuing to stack up procedure-description capabilities.
All Facts (2)
BEAKER's task types include affordance localization, trajectory planning, equipment/state/safety recognition, apparatus setup, operation and step reasoning, and temporal procedure reasoning.
50%UnverifiedFuture model improvements for laboratories should prioritize filling in "action grounding" capabilities such as affordance localization and trajectory planning, rather than continuing to stack up procedure-description capabilities.
50%