Self-Information Loss Compensation Learning for Machine-Generated Text Detection

Table 3

Comparison of results with the F1 score and accuracy.

Sequence length	64		128		256		512
Models	F1 score	ACC	F1 score	ACC	F1 score	ACC	F1 score	ACC
FastText	0.7775	0.8131	0.7861	0.8200	0.7853	0.8169	0.7822	0.8125
TextCNN	0.7811	0.8067	0.7848	0.8131	0.7905	0.8183	0.7858	0.8128
LSTM	0.8414	0.8544	0.7907	0.8219	0.5923	0.6811	0.0044	0.5003
RCNN	0.8570	0.8650	0.8982	0.9033	0.8977	0.9036	0.9021	0.9061
LSTM + Attention	0.8547	0.8633	0.8841	0.8872	0.8770	0.8819	0.8844	0.8889
Our model	0.8731	0.8792	0.9072	0.9103	0.9003	0.9050	0.9131	0.9158