A baseline TGM model which evaluates the log likelihood probability, and thresholds based on this probability to make the prediction.

Total log probability for TGMs

A suite of statistical methods which can highlight the differences in text generated by GPT-2 model and human written text.

Giant Language model Test Room (GLTR) tool

In this a pre trained TGM model is used to classify texts generated from its model or similar models.

 - Total log probability
 - Giant Language model Test Room (GLTR) tool


San Jose State University

Detectors are used to identify machine generated text from human generated text. Classification of detectors are based on

- Classifiers trained from scratch 
 - Zero-shot classifier 
 - Fine-tuning NLM 
 - Human-machine collaboration


TGM Classifiers or Detectors

Automatic Detection of Machine Generated Text: A Critical Survey
by Ganesh Jawahar, Muhammad Abdul-Mageed, Laks V.S. Lakshmanan
https://arxiv.org/pdf/2011.01314.pdf

Automatic Detection of Machine Generated Text: A Critical Survey


 - Bag-of-words classifier
 - Detecting machine configuration


Classifiers trained from scratch for machine generated text detection

Zero-shot classifier for TGMs

Fine tuning pretrained models to detect text samples from its own model or other similar models .the fine tuned pretrained model doesn’t require further supervised example training.

 - GROVER detector
 - RoBERTa detector


Learn Before

Related

Learn After