Shauli Ravfogel

A new method to improve language model behavior by reducing bias and toxic output.

Discover how language imbalance can improve multilingual model performance.

New methods reveal challenges in unlearning knowledge from language models.

Research shows diversity in training data is key for better model performance.