← All threads

Object Permanence in VLMs

Benchmarks and methods for evaluating vision-language models' ability to track objects after they leave the field of view.

3 papers

Where this stands

The written synthesis of this thread is for subscribers. Subscribe.