Radar Tester

Robust AI-Text Detection via Adversarial Learning

Radar tester is based on the paper "RADAR: Robust AI-Text Detection via Adversarial Learning" by Xiaomeng Hu (CUHK), Pin-Yu Chen (IBM) , Tsung-Yi Ho (CUHK).
This demo was created by Hendrik Strobelt (IBM, MIT) and Ben Hoover (IBM, GATech) .

For technical details, please visit the project page

Select an examples (or enter your test below):

Please enter some text to check

Should be at least three words.

Select test models...

A contact method is required

Results

For each text we show how likely each model thinks, the text is generated by AI. A value close to 1 indicates "most likely AI", a value close to 0 means "most likely human". When the models show very different results, it's most likely human as well.