Skip to Main Content
Table 2: 
Micro-averaged accuracy of different methods (%). Majority gives us 22.0%. Italic indicates best single-prompt accuracy, and bold indicates the best non-oracle accuracy overall.
PromptsTop1Top3Top5Opti.Oracle
BERT-base (Man=31.1) 
Mine 31.4 34.2 34.7 38.9 50.7 
Mine+Man 31.6 35.9 35.1 39.6 52.6 
Mine+Para 32.7 34.0 34.5 36.2 48.1 
Man+Para 34.1 35.8 36.6 37.3 47.9 
 
BERT-large (Man=32.3) 
Mine 37.0 37.0 36.4 43.7 54.4 
Mine+Man 39.4 40.6 38.4 43.9 56.1 
Mine+Para 37.8 38.6 38.6 40.1 51.8 
Man+Para 35.9 37.3 38.0 38.8 50.0 
PromptsTop1Top3Top5Opti.Oracle
BERT-base (Man=31.1) 
Mine 31.4 34.2 34.7 38.9 50.7 
Mine+Man 31.6 35.9 35.1 39.6 52.6 
Mine+Para 32.7 34.0 34.5 36.2 48.1 
Man+Para 34.1 35.8 36.6 37.3 47.9 
 
BERT-large (Man=32.3) 
Mine 37.0 37.0 36.4 43.7 54.4 
Mine+Man 39.4 40.6 38.4 43.9 56.1 
Mine+Para 37.8 38.6 38.6 40.1 51.8 
Man+Para 35.9 37.3 38.0 38.8 50.0 
Close Modal

or Create an Account

Close Modal
Close Modal