“My Answer is C”: First-Token Probabilities Do Not Match Text Answers in Instruction-Tuned Language Models Xinpeng Wang author Bolei Ma author Chengzhi Hu author Leon Weber-Genzel author Paul Röttger author Frauke Kreuter author Dirk Hovy author Barbara Plank author 2024-08 text Findings of the Association for Computational Linguistics: ACL 2024 Lun-Wei Ku editor Andre Martins editor Vivek Srikumar editor Association for Computational Linguistics Bangkok, Thailand conference publication wang-etal-2024-answer-c 10.18653/v1/2024.findings-acl.441 https://aclanthology.org/2024.findings-acl.441/ 2024-08 7407 7416