(she/her/hers)
University of Washington and Meta AI
Natural Language Processing, Multilingual Modeling, Ethics, Interpretation
Hila is a postdoctoral researcher at Meta AI and at the Paul G. Allen School of Computer Science & Engineering at the University of Washington. Her research focuses on multilingual NLP, both from the perspective of modeling and from the perspective of analysis. She also works on promoting fairness in NLP, focusing mainly on gender bias identification and mitigation.
Hila is the recipient of three prestigious postdoctoral fellowships/awards and two best paper awards (Rothschild, Fulbright, the Eric and Wendy Schmidt Postdoctoral Award for women in CS). Before joining UW and Meta AI, Hila was a postdoctoral researcher at Amazon. Prior to that she did her Ph.D in Computer Science at the NLP lab at Bar Ilan University. She obtained her Ms.C. in Computer Science from the Hebrew University.
Can large language models translate and how?
A lot of the progress in Natural Language Processing today is based on large language models that are trained with a lot of text with no specific downstream task at hand. These models are very powerful and allow classification and generation even without further training for a specific task but just with prompting the model: we simply input an instruction or a relevant question or sentence to get the appropriate answer from the model. While a lot of progress is made in English, less attention is given to modeling languages outside of English. However, when we turn to multilingual modeling, many interesting phenomena are revealed, allowing the model to transfer information across languages. In this talk I will focus on the unique characteristics of these multilingual models, and provide some explanations to their astonishing capabilities.