Skip to Main Content
Table 8: 
Accuracy (Acc, %) of relation ranking and winning rates (%) of pairwise comparisons which require selecting a more reasonable sentence from two candidates, each from Correct (C), Wrong (W), or Training (T) templates.
Comparison Pairs
ModelsAccC vs. WT vs. CT vs. W
GPT-2 (Pretrain) 39.28 53.83 44.74 49.87 
GPT-2(Fine-tune) 47.48 60.31 39.57 56.01 
 
Ours 67.07 71.91 55.76 79.89 
w/o Knowledge 48.07 62.07 42.43 55.64 
Comparison Pairs
ModelsAccC vs. WT vs. CT vs. W
GPT-2 (Pretrain) 39.28 53.83 44.74 49.87 
GPT-2(Fine-tune) 47.48 60.31 39.57 56.01 
 
Ours 67.07 71.91 55.76 79.89 
w/o Knowledge 48.07 62.07 42.43 55.64 
Close Modal

or Create an Account

Close Modal
Close Modal