Skip to Main Content
Table 5

Performance of different post-hoc methods using the UnifiedQA model after margin-based fine-tuning or the original UnifiedQA model as the baseline model. “+Combo” denotes the method using both Temp., Para., and Aug.

MethodMC-testMT-testExt-test
ACCECEACCECEACCECE
Baseline 0.769 0.057 0.431 0.144 0.401 0.114 
+ Temp. 0.769 0.049 0.431 0.075 0.401 0.107 
+ XGB 0.771 0.055 0.431 0.088 0.402 0.103 
+ Para. 0.767 0.051 0.429 0.122 0.393 0.114 
+ Aug. 0.744 0.051 0.432 0.130 0.408 0.110 
+ Combo 0.748 0.044 0.431 0.079 0.398 0.104 
MethodMC-testMT-testExt-test
ACCECEACCECEACCECE
Baseline 0.769 0.057 0.431 0.144 0.401 0.114 
+ Temp. 0.769 0.049 0.431 0.075 0.401 0.107 
+ XGB 0.771 0.055 0.431 0.088 0.402 0.103 
+ Para. 0.767 0.051 0.429 0.122 0.393 0.114 
+ Aug. 0.744 0.051 0.432 0.130 0.408 0.110 
+ Combo 0.748 0.044 0.431 0.079 0.398 0.104 
Close Modal

or Create an Account

Close Modal
Close Modal