I-CARE: A New Framework to Measure Interference in Machine Unlearning for Text-to-Image Models

Machine unlearning is the process of removing knowledge from a trained AI model, making it forget concepts it previously learned.

Interference is the unintended degradation of semantically related concepts that should have been retained when a concept is removed.

I-CARE provides a standardized methodology to evaluate interference risks, helping organizations make informed decisions about using text-to-image models.

Yes, it is available as an open-source framework with a web-based GUI, making it accessible without deep programming expertise.