VLMs with harness have recently become very competitive with VLAs see for example
https://showlab.github.io/Show-Harness/
The question is how to adapt your method to a Vlm based harness ? Since it has no action heads I wonder how to use your solution..
VLMs with harness have recently become very competitive with VLAs see for example
https://showlab.github.io/Show-Harness/
The question is how to adapt your method to a Vlm based harness ? Since it has no action heads I wonder how to use your solution..