On the Calibration of Generative Question-Answering models: State-of-the-art and Challenges

Nowadays, generative question-answering models (e.g., UnifiedQA) achieve state-of-the-art performance in various datasets. Despite their remarkable performance, these models still produce wrong answers with high confidence scores. The responsible use of such systems in high-risk applications, like healthcare, requires some guarantees in terms of the correlation of the model scores and the output’s correctness. One potential approach toward these guarantees is calibration. Despite the vast research on calibration for K-ary classification in machine learning, calibration for textual-based systems imposes additional challenges that range from a combinatorial output space to the many definitions of correctness. In this talk, we will discuss the challenges towards the calibration of generative question-answering systems, as well as the current state-of-the-art approaches to address it.

Catarina Belém

Catarina Belém is a first-year Ph.D. candidate in Computer Science at the University of California Irvine (UCI). Currently, she is working under the supervision of professors Sameer Singh and Padhraic Smyth on the calibration of generative question-answering models. Prior to joining UCI, Catarina worked as a research data scientist at the Responsible AI (FATE) group at Feedzai, where she developed a keen interest in fairness, explainability, and evaluation in AI. Catarina holds an integrated master’s degree (BSc+MSc) in Computer Engineering obtained from Instituto Superior Tecnico in 2019. Her main research interests include Machine Learning and Natural Language Processing with a particular focus on Responsible AI.University of California Irvine