“My Answer is C”: First-Token Probabilities Do Not Match Text Answers in Instruction-Tuned Language Models

Xinpeng Wang, Bolei Ma, Chengzhi Hu, Leon Weber-Genzel, Paul Röttger, Frauke Kreuter, Dirk Hovy, Barbara Plank


Abstract
The open-ended nature of language generation makes the evaluation of autoregressive large language models (LLMs) challenging. One common evaluation approach uses multiple-choice questions to limit the response space. The model is then evaluated by ranking the candidate answers by the log probability of the first token prediction. However, first-tokens may not consistently reflect the final response output, due to model’s diverse response styles such as starting with “Sure” or refusing to answer. Consequently, first-token evaluation is not indicative of model behaviour when interacting with users. But by how much? We evaluate how aligned first-token evaluation is with the text output along several dimensions, namely final option choice, refusal rate, choice distribution and robustness under prompt perturbation. Our results show that the two approaches are severely misaligned on all dimensions, reaching mismatch rates over 60%. Models heavily fine-tuned on conversational or safety data are especially impacted. Crucially, models remain misaligned even when we increasingly constrain prompts, i.e., force them to start with an option letter or example template. Our findings i) underscore the importance of inspecting the text output as well and ii) caution against relying solely on first-token evaluation.
Anthology ID:
2024.findings-acl.441
Volume:
Findings of the Association for Computational Linguistics: ACL 2024
Month:
August
Year:
2024
Address:
Bangkok, Thailand
Editors:
Lun-Wei Ku, Andre Martins, Vivek Srikumar
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
7407–7416
Language:
URL:
https://aclanthology.org/2024.findings-acl.441
DOI:
10.18653/v1/2024.findings-acl.441
Bibkey:
Cite (ACL):
Xinpeng Wang, Bolei Ma, Chengzhi Hu, Leon Weber-Genzel, Paul Röttger, Frauke Kreuter, Dirk Hovy, and Barbara Plank. 2024. “My Answer is C”: First-Token Probabilities Do Not Match Text Answers in Instruction-Tuned Language Models. In Findings of the Association for Computational Linguistics: ACL 2024, pages 7407–7416, Bangkok, Thailand. Association for Computational Linguistics.
Cite (Informal):
“My Answer is C”: First-Token Probabilities Do Not Match Text Answers in Instruction-Tuned Language Models (Wang et al., Findings 2024)
Copy Citation:
PDF:
https://preview.aclanthology.org/landing_page/2024.findings-acl.441.pdf