A benchmark for toxic comment classification on Civil Comments dataset

Benchmark Model Rank Results
toxic-comment-classification-on-civilRoBERTa Focal Loss#1GMB BPSN: 0.901GMB Subgroup: 0.8807Micro F1: 0.5524
toxic-comment-classification-on-civilAlBERT#2GMB BPSN: 0.8982GMB Subgroup: 0.8734Micro F1: 0.4845
toxic-comment-classification-on-civilBERTweet#3GMB BPSN: 0.8945GMB Subgroup: 0.878Micro F1: 0.4928
toxic-comment-classification-on-civilHateBERT#4GMB BPSN: 0.8915GMB Subgroup: 0.8744Micro F1: 0.4844
toxic-comment-classification-on-civilRoBERTa BCE#5GMB BPSN: 0.8901GMB Subgroup: 0.88Micro F1: 0.5359
toxic-comment-classification-on-civilXLM RoBERTa#6GMB BPSN: 0.8859Micro F1: 0.468Precision: 0.3135Recall: 0.923
toxic-comment-classification-on-civilXLNet#7GMB BPSN: 0.8834GMB Subgroup: 0.8689Micro F1: 0.4586
toxic-comment-classification-on-civilDistilBERT#8GMB BPSN: 0.874GMB Subgroup: 0.8762Micro F1: 0.5115
toxic-comment-classification-on-civilBiGRU#9GMB BPSN: 0.8616
toxic-comment-classification-on-civilUnfreeze Glove ResNet 44#10GMB BPSN: 0.8493GMB Subgroup: 0.8421Micro F1: 0.5958
toxic-comment-classification-on-civilUnfreeze Glove ResNet 56#11GMB BPSN: 0.8445GMB Subgroup: 0.8487Recall: 0.8707Macro F1: 0.3778
toxic-comment-classification-on-civilCompact Convolutional Transformer (CCT)#12GMB BPSN: 0.8307GMB Subgroup: 0.8133Micro F1: 0.4874
toxic-comment-classification-on-civilFreeze Glove ResNet 44#13GMB BPSN: 0.7876GMB Subgroup: 0.8219Micro F1: 0.5591
toxic-comment-classification-on-civilBiLSTM#14GMB Subgroup: 0.8636Micro F1: 0.5115Precision: 0.3572