Publications

Selected Publications feb 2024 "My Answer is C": First-Token Probabilities Do Not Match Text Answers in Instruction-Tuned Language Models This paper investigates to what extent the first token probabilities of large language models match their final answers to multiple-choice questions. dec 2023 Establishing Trustworthiness: Rethinking Tasks and Model Evaluation Decades of NLP research have traditionally compartmentalized linguistic tasks, but the emergence of large language models is reshaping this approach, emphasizing the need for holistic, task-agnostic evaluation methods that prioritize trustworthiness. may 2023 A Survey of Corpora for Germanic Low-Resource Languages and Dialects This paper provides an overview of more than 80 corpora to support NLP research in resource-poor and non-standardized languages of the Germanic language family.