OpenGVLab/LAMM

What are the metrics of the six recipes of Desiderata?

zhimin-z opened this issue · 4 comments

image
I fail to find any of those recipes in the original paper...

BTW, are those metrics all accuracy?

Desiderata is a newly proposed evaluation metric in ChEF, focusing on dimensions of capabilities beyond the visual abilities of MLLMs. These metrics include the trustworthiness and interactivity. Please refer to our paper ChEF for more details. Thanks for your interest.

image
But I still fail to find this one in terms of the exact evaluation metrics. Is the evaluation result accuracy or not?

image But I still fail to find this one in terms of the exact evaluation metrics. Is the evaluation result accuracy or not?

Any update? @Coach257