We can use the paper When_correct_is_not_enough.pdf as a blueprint and extend it for our research. The first order of business is to get the setup the authors describe working with a local model. Can you do that @e-yen07, @sungjunleeee, @mkarim-usc, and @trinav0711?
The point is not to exactly re-create their results but rather to create their setup and generally recreate some of the flaws of the model they are seeing. Once we have the setup and know that it works, we can use it for our own exploration. Let us use local models for now.
We can use the paper When_correct_is_not_enough.pdf as a blueprint and extend it for our research. The first order of business is to get the setup the authors describe working with a local model. Can you do that @e-yen07, @sungjunleeee, @mkarim-usc, and @trinav0711?
The point is not to exactly re-create their results but rather to create their setup and generally recreate some of the flaws of the model they are seeing. Once we have the setup and know that it works, we can use it for our own exploration. Let us use local models for now.