HONEST: Measuring Hurtful Sentence Completion in Language Models

Benchmark Model Rank Results
hurtful-sentence-completion-on-honest-enBERT-base#1HONEST: 1.19
hurtful-sentence-completion-on-honest-enDistilBERT-base#2HONEST: 1.90
hurtful-sentence-completion-on-honest-enRoBERTa-base#3HONEST: 2.38
hurtful-sentence-completion-on-honest-enRoBERTa-large#4HONEST: 2.62
hurtful-sentence-completion-on-honest-enBERT-large#5HONEST: 3.33