What do Models Learn From Training on More Than Text? Measuring Visual Commonsense Knowledge

Lovisa Hagström, Richard Johansson


Abstract
There are limitations in learning language from text alone. Therefore, recent focus has been on developing multimodal models. However, few benchmarks exist that can measure what language models learn about language from multimodal training. We hypothesize that training on a visual modality should improve on the visual commonsense knowledge in language models. Therefore, we introduce two evaluation tasks for measuring visual commonsense knowledge in language models (code publicly available at: github.com/lovhag/measure-visual-commonsense-knowledge) and use them to evaluate different multimodal models and unimodal baselines. Primarily, we find that the visual commonsense knowledge is not significantly different between the multimodal models and unimodal baseline models trained on visual text data.
Anthology ID:
2022.acl-srw.19
Volume:
Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop
Month:
May
Year:
2022
Address:
Dublin, Ireland
Editors:
Samuel Louvan, Andrea Madotto, Brielen Madureira
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
252–261
Language:
URL:
https://aclanthology.org/2022.acl-srw.19
DOI:
10.18653/v1/2022.acl-srw.19
Bibkey:
Cite (ACL):
Lovisa Hagström and Richard Johansson. 2022. What do Models Learn From Training on More Than Text? Measuring Visual Commonsense Knowledge. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop, pages 252–261, Dublin, Ireland. Association for Computational Linguistics.
Cite (Informal):
What do Models Learn From Training on More Than Text? Measuring Visual Commonsense Knowledge (Hagström & Johansson, ACL 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.acl-srw.19.pdf
Video:
 https://aclanthology.org/2022.acl-srw.19.mp4
Code
 lovhag/measure-visual-commonsense-knowledge