A big thank you to Pushkar Shukla , Ph.D. student at Toyota Technological Institute at Chicago , for delivering an insightful talk on 14th January 2025.
Pushkar explored the biases in Text-to-Image (TTI) generative models, including societal stereotypes and incidental correlations, and introduced TIBET, a novel approach for evaluating and quantifying these biases using counterfactual reasoning. He also discussed diagnosing intersectionality in biases and the challenges of creating fair and diverse generative systems.
We are also grateful to Prof. Saket Anand for connecting us with Pushkar and making this enlightening session possible at Indraprastha Institute of Information Technology, Delhi .