Chrysoula Zerva

Chrysoula Zerva is an Invited Assistant Professor in the Department of Electrical and Computer Engineering at Instituto Superior Técnico and a post-doc researcher in the SARDINE group at the Instituto de Telecomunicações. She holds a PhD from the University of Manchester (2019) and was awarded the EPSRC Doctoral Prize Fellowship for the “Fake Health News” project. Her research focuses on uncertainty quantification, fairness, explainability in NLP, and machine translation evaluation. Key interests include adversarial attacks, bias mitigation, and disentangled representation learning.

Dr. Zerva’s work spans conformal prediction frameworks, dialect bias in reward models, and response clarity classification. Her contributions to MT evaluation and quality estimation have been highlighted in WMT shared tasks. Recent studies address ethical AI challenges, such as the societal impacts of LLMs and biases against African American language.

• Scientific Awards: EPSRC Doctoral Prize Fellowship
• Labs/Teams: Member of the SARDINE research group

Her publications investigate cutting-edge topics like CLIPScore metrics, non-exchangeable conformal methods, and context-aware NMT for business dialogues. Collaborations include Unbabel and participation in international shared tasks.