Multilingual Performance of a Multimodal Artificial Intelligence System on Multisubject Physics Concept Inventories
Gerd Kortemeyer 1,2 , Marina Babayeva 3 , Giulia Polverini 4 , Ralf Widenhorn 5 , Bor Gregorcic 4 · 2025
Read Multilingual Performance of a Multimodal Artificial Intelligence System on Multisubject Physics Concept Inventories by Gerd Kortemeyer 1,2 , Marina Babayeva 3 , Giulia Polverini 4 , Ralf Widenhorn 5 , Bor Gregorcic 4 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-06143.txt.
Description
Gerd Kortemeyer and co-authors test in 2025 how a multimodal AI system handles physics concept tests in many languages and subjects. Concept inventories are standard surveys that probe student grasp of ideas in mechanics, electricity, optics, heat, and more advanced fields. Past studies fed the system plain text only.
Here the tests were uploaded as images, as a student would see them on paper, drawn from a public physics teaching site. Scores varied by subject, with laboratory skills the weakest area, and by language, with English and other European languages strongest. Which items were hard barely changed across languages. Against published student data, the system beat average students after teaching in every subject but laboratory skills. Items needing image reading tripped it more than pure text items, a caution for classroom use.
Catalog record
| CREATOR | Gerd Kortemeyer 1,2 , Marina Babayeva 3 , Giulia Polverini 4 , Ralf Widenhorn 5 , Bor Gregorcic 4 |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 90 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-06143