Describe the bug
The evaluation script and configurations for AdvGlue use only 1 template that is fixed and is not configurable in the run configs. In the paper, you state that you computed the adv performance using 3 different templates.
To Reproduce
Not Relevant
Expected behavior
I would expect the code to support a 'template' parameter for each advglue config to enable performance evaluation with different templates like stated in the paper
Environment:
Not Relevant
Additional context
Not Relevant
Describe the bug
The evaluation script and configurations for AdvGlue use only 1 template that is fixed and is not configurable in the run configs. In the paper, you state that you computed the adv performance using 3 different templates.
To Reproduce
Not Relevant
Expected behavior
I would expect the code to support a 'template' parameter for each advglue config to enable performance evaluation with different templates like stated in the paper
Environment:
Not Relevant
Additional context
Not Relevant