Toxic Comment Classification On Civil

Metrics

AUROC

Results

Performance results of various models on this benchmark

Model Name	AUROC	Paper Title
LightGBM + RoBERTa embedding	0.865	PyTorch Frame: A Modular Framework for Multi-Modal Tabular Learning
BiLSTM	-	A benchmark for toxic comment classification on Civil Comments dataset
Unfreeze Glove ResNet 44	0.966	A benchmark for toxic comment classification on Civil Comments dataset
Compact Convolutional Transformer (CCT)	0.9526	A benchmark for toxic comment classification on Civil Comments dataset
BiGRU	-	A benchmark for toxic comment classification on Civil Comments dataset
Freeze Glove ResNet 44	-	A benchmark for toxic comment classification on Civil Comments dataset
BERTweet	0.979	A benchmark for toxic comment classification on Civil Comments dataset
XLNet	-	A benchmark for toxic comment classification on Civil Comments dataset
ResNet + RoBERTa embedding	0.882	PyTorch Frame: A Modular Framework for Multi-Modal Tabular Learning
Trompt + OpenAI embedding	0.947	PyTorch Frame: A Modular Framework for Multi-Modal Tabular Learning
XLM RoBERTa	-	A benchmark for toxic comment classification on Civil Comments dataset
DistilBERT	0.9804	A benchmark for toxic comment classification on Civil Comments dataset
PaLM 2 (zero-shot)	0.7596	PaLM 2 Technical Report
ResNet + RoBERTa finetune	0.97	PyTorch Frame: A Modular Framework for Multi-Modal Tabular Learning
RoBERTa Focal Loss	0.9818	A benchmark for toxic comment classification on Civil Comments dataset
RoBERTa BCE	0.9813	A benchmark for toxic comment classification on Civil Comments dataset
PaLM 2 (few-shot, k=10)	0.8535	PaLM 2 Technical Report
Unfreeze Glove ResNet 56	0.9639	A benchmark for toxic comment classification on Civil Comments dataset
HateBERT	0.9791	A benchmark for toxic comment classification on Civil Comments dataset
ResNet + OpenAI embedding	0.945	PyTorch Frame: A Modular Framework for Multi-Modal Tabular Learning

0 of 22 row(s) selected.