Skip to Main Content
Table 7: 
Test results for M + C* method on different permutations of modified-bAbI dialog task’s test set.
User AccuracyModel ratio
Per-turnper-dialog
Baseline method (M) 
81.73 3.7 100.0 
 
Reward: 1, 2, -4, (M + C*
92.44 33.1 50.98 
92.57 32.7 53.22 
93.96 42.0 44.63 
94.7 43.0 45.82 
90.59 16.6 65.24 
 
Reward: 1, 3, -3, (M + C*
90.41 21.1 54.67 
92.25 36.1 54.79 
92.12 32.5 58.07 
89.75 18.2 65.69 
92.05 24.6 60.92 
User AccuracyModel ratio
Per-turnper-dialog
Baseline method (M) 
81.73 3.7 100.0 
 
Reward: 1, 2, -4, (M + C*
92.44 33.1 50.98 
92.57 32.7 53.22 
93.96 42.0 44.63 
94.7 43.0 45.82 
90.59 16.6 65.24 
 
Reward: 1, 3, -3, (M + C*
90.41 21.1 54.67 
92.25 36.1 54.79 
92.12 32.5 58.07 
89.75 18.2 65.69 
92.05 24.6 60.92 
Close Modal

or Create an Account

Close Modal
Close Modal