Antoine Saillenfest

New MUtE framework enhances AI fairness and text generation

Researchers have developed MUtE, a dual framework for concept erasure and counterfactual interventions in language models. This framework aims to remove concept-specific information from model representations while pres…