Benchmark

Find-the-Item

A real-world robotic object-finding task that assesses a model's ability to remember and make use of past observations.

Timeline (last 90 days)

Oct 21

In the real-world Find-the-Item task, generated textual memory raises UniWALM's success rate from 11.1% to 61.1%.

Unverified50%

Oct 21

On the Find-the-Item task, current state-of-the-art VLA models are only at or below chance level (success rate ≤33.3%).

Unverified50%

In the real-world Find-the-Item task, generated textual memory raises UniWALM's success rate from 11.1% to 61.1%.

On the Find-the-Item task, current state-of-the-art VLA models are only at or below chance level (success rate ≤33.3%).