Every Answer Matters: Evaluating Commonsense with Probabilistic Measures

Qi Cheng, Michael Boratko, Pranay Kumar Yelugam, Tim O’Gorman, Nalini Singh, Andrew McCallum, Xiang Li


Abstract
Large language models have demonstrated impressive performance on commonsense tasks; however, these tasks are often posed as multiple-choice questions, allowing models to exploit systematic biases. Commonsense is also inherently probabilistic with multiple correct answers. The purpose of “boiling water” could be making tea, cooking but also could be killing germs. Existing tasks do not capture the probabilistic nature of common sense. To this end, we present commonsense frame completion (CFC), a new generative task that evaluates common sense via multiple open-ended generations. We also propose a method of probabilistic evaluation that strongly correlates with human judgments. Humans drastically outperform strong language model baselines on our dataset, indicating this approach is both a challenging and useful evaluation of machine common sense.
Anthology ID:
2024.acl-long.29
Volume:
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
August
Year:
2024
Address:
Bangkok, Thailand
Editors:
Lun-Wei Ku, Andre Martins, Vivek Srikumar
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
493–506
Language:
URL:
https://aclanthology.org/2024.acl-long.29
DOI:
10.18653/v1/2024.acl-long.29
Bibkey:
Cite (ACL):
Qi Cheng, Michael Boratko, Pranay Kumar Yelugam, Tim O’Gorman, Nalini Singh, Andrew McCallum, and Xiang Li. 2024. Every Answer Matters: Evaluating Commonsense with Probabilistic Measures. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 493–506, Bangkok, Thailand. Association for Computational Linguistics.
Cite (Informal):
Every Answer Matters: Evaluating Commonsense with Probabilistic Measures (Cheng et al., ACL 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.acl-long.29.pdf