Multilingual Performance of a Multimodal Artificial Intelligence System on Multisubject Physics Concept Inventories Documents
You are not currently logged in so we can not check your permission status. Please login in order to see if you may access this content.
Primary Documents (2)
Dataset and Public Code for "Multilingual Performance of a Multimodal Artificial Intelligence System on Multisubject Physics Concept Inventories"
by Gerd Kortemeyer, Marina Babayeva, Giulia Polverini, Ralf Widenhorn, and Bor Gregorcic
This dataset accompanies the manuscript: "Multilingual Performance of a Multimodal Artificial Intelligence System on Multisubject Physics Concept Inventories."
This package is freely accessible and contains most of the analysis code:
* Analysis - CSV files used in the analysis (except for a "Normed Answer" column)
* GPT4oResults - this directory is redacted, but information for how researchers may access the data is provided in an Access.txt file.
* InventoryImages this directory is redacted, but information for how researchers may access the data is provided in an Access.txt file.
* Programs - some of the Python code used
This file should allow researchers the ability to look at the code analyzing the assessments.
If you are an AI researcher that wishes to run the analysis yourself, you may gain access to the unredacted files. Please create an account on PER-Central.org or PhysPort.org and provide verification@physport.org with your account name and research credentials using the subject line: "Access to Kortemeyer et al. AI & Concept Inventory Compendium"
After verification, your account will receive permission to download the file and you will receive a password that you may use to open the file.
- Download ConceptInventoryAIStudy - Public.zip - 50kb Compressed File
Published June 7, 2025
Last Modified June 7, 2025
This file has previous versions.
Confidential Dataset and Code for "Multilingual Performance of a Multimodal Artificial Intelligence System on Multisubject Physics Concept Inventories"
Verified Researchers
If you have an account with permission to access this file, please login to access this restricted file.
by Gerd Kortemeyer, Marina Babayeva, Giulia Polverini, Ralf Widenhorn, and Bor Gregorcic
| Type: | Primary |
|---|---|
| Title: | Confidential Dataset and Code for "Multilingual Performance of a Multimodal Artificial Intelligence System on Multisubject Physics Concept Inventories" |
| Credits: | Gerd Kortemeyer, Marina Babayeva, Giulia Polverini, Ralf Widenhorn, and Bor Gregorcic |
| Description: | This dataset accompanies the manuscript: Multilingual Performance of a Multimodal Artificial Intelligence System on Multisubject Physics Concept Inventories by Gerd Kortemeyer, Marina Babayeva, Giulia Polverini, Ralf Widenhorn, and Bor Gregorcic. The dataset is CONFIDENTIAL; it needs to be treated under the same rules and guidelines as the assessments in PhysPort in order to not compromise their research validity. This package contains: * GPT4oResults - JSON files with the LLM outputs, as discussed in the appendix of the manuscript * Analysis - CSV files used in the analysis * Programs - some of the Python code used * InventoryImages - screenshots of the inventory questions, as discussed in the manuscript For someone to gain access (even collaborators), please have them create an account on PER-Central or PhysPort and then provide their research credentials to verification@physport.org with the subject line: "Access to Kortemeyer et al. AI & Concept Inventory Compendium". Their account will receive the ability to download this file and they will receive a password that will allow them to open the file. |
| Access Restricted To: | Verified Researchers |
| Formats: |
Available formats: |
| Rights *: | This dataset is confidential and may not be shared either publicly or with collaborators. If a collaborator needs access, please have them create an account on PER-Central or PhysPort and then provide their research credentials to verification@physport.org with the subject line: "Access to Kortemeyer et al. AI & Concept Inventory Compendium". Their account will receive the ability to download this file and they will receive a password that will allow them to open the file. |
| Publication Date: | June 7, 2025 |
| Created: | June 7, 2025 at 21:10:00 EST |
| Version Notes: | Document set created. |
* These were the rights specified when this version of the document was created. For new downloads, the rights for the document are those attributed to the most recent version.
