top of page

EvoGrad: Help fool (and improve) machine learning models!

Evograd is an open source platform for the continual evaluation
and development of models, based on iterations of human-adversarial perturbations. Inspired by the Winograd Schema Challenge, this task is designed to test the common-sense capabilities of deep learning models using simple coreference resolution problems. 

Read Paper:

download.png
dynagrad.jpeg
bottom of page