Table 9: 
Hyper-parameter tuning.
TaskParamSearch SpaceSelected Value
Open Domain (Local) Learning Rate 2e-6,5e-6,2e-5,5e-5 2e-5 
Batch size 32 (Max. Mem) 32 
Patience 1,3,5 3 
Optimizer SGD,Adam,AdamW AdamW 
Hidden Units 128,512 512 
Non-linearity − ReLU 
 
Open Domain (Global) Learning Rate 2e-6,5e-6,2e-5,5e-5 2e-6 
Batch size − Full instance 
 
Stance Pred. (Local) Learning Rate 2e-6,5e-6,2e-5,5e-5 5e-5 
Patience 1,3,5 3 
Batch size 16 (Max. Mem) 16 
Optimizer SGD,Adam,AdamW AdamW 
 
Stance Pred. (Global) Learning Rate 2e-6,5e-6,2e-5,5e-5 2e-6 
Batch size − Full instance 
 
Arg. Mining (Local) Learning Rate 1e-4,5e-4,5e-3,1e-3,5e-2,1e-2 5e-2 
Patience 5,10,20 20 
Batch size 16,32,64,128 64 
Dropout 0.01,0.05,0.1 0.05 
Optimizer SGD,Adam,AdamW SGD 
Hidden Units 128,512 128 
Non-linearity − ReLU 
 
Arg. Mining (Global) Learning Rate 1e-4,5e-4,5e-3,1e-3,5e-2,1e-2 1e-4 
Patience 5,10,20 10 
Batch size − Full instance 
TaskParamSearch SpaceSelected Value
Open Domain (Local) Learning Rate 2e-6,5e-6,2e-5,5e-5 2e-5 
Batch size 32 (Max. Mem) 32 
Patience 1,3,5 3 
Optimizer SGD,Adam,AdamW AdamW 
Hidden Units 128,512 512 
Non-linearity − ReLU 
 
Open Domain (Global) Learning Rate 2e-6,5e-6,2e-5,5e-5 2e-6 
Batch size − Full instance 
 
Stance Pred. (Local) Learning Rate 2e-6,5e-6,2e-5,5e-5 5e-5 
Patience 1,3,5 3 
Batch size 16 (Max. Mem) 16 
Optimizer SGD,Adam,AdamW AdamW 
 
Stance Pred. (Global) Learning Rate 2e-6,5e-6,2e-5,5e-5 2e-6 
Batch size − Full instance 
 
Arg. Mining (Local) Learning Rate 1e-4,5e-4,5e-3,1e-3,5e-2,1e-2 5e-2 
Patience 5,10,20 20 
Batch size 16,32,64,128 64 
Dropout 0.01,0.05,0.1 0.05 
Optimizer SGD,Adam,AdamW SGD 
Hidden Units 128,512 128 
Non-linearity − ReLU 
 
Arg. Mining (Global) Learning Rate 1e-4,5e-4,5e-3,1e-3,5e-2,1e-2 1e-4 
Patience 5,10,20 10 
Batch size − Full instance 
View Large
Close Modal
This Feature Is Available To Subscribers Only

or Create an Account

Close Modal
Close Modal
This site uses cookies. By continuing to use our website, you are agreeing to our privacy policy. No content on this site may be used to train artificial intelligence systems without permission in writing from the MIT Press.
Accept