ScandEval icon indicating copy to clipboard operation
ScandEval copied to clipboard
verified publisher icon ScandEval

Truthfulness evaluation

Open saattrupdan opened this issue 1 year ago • 1 comments

This will add an orthogonal evaluation of decoder language models, testing how much they hallucinate.

saattrupdan avatar Jan 20 '25 17:01 saattrupdan

This dataset might be relevant: https://openai.com/index/introducing-simpleqa/

saattrupdan avatar Jan 20 '25 17:01 saattrupdan