NLI
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
NLI shared evaluation pool
3,277.7Mean Cost-Weighted Pseudo-Regret
18
NLI adversarial benchmark (test)
75.4Average Score
18
NLI
91.2Accuracy
14
NLI ANLI and HANS (unseen)
32.4ANLI Score
9
NLI
100Detection Rate (TPR/FPR)
8
NLI domain average
87.5Best Accuracy
8
NLI
97.9RL Score
3
NLI (test)
2.89Relative CPU Speed
2