Doodling AI: CrowS-Pairs Challenge

Starting a new series of posts where I doodle a gentle overview of ML/AI research papers that I find interesting. If you find it interesting, please refer to the paper for more details.

Today’s Topic: CrowS-Pairs Challenge Dataset

With the growing popularity of pre-trained language models, it is important to understand their biases before their real-world applications. These studies will not only encourage reserach to remove these biases, but also give insights for regulating omnipresent AI tools.

Last Words: Let me know your thoughts in the comments. Constructive criticism is most welcome!

References
[1] CrowS-Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Models https://arxiv.org/abs/2010.00133