This model is the outcome of a summer internship carried out at Max Planck's Center for Humans and Machines group. The methodology and ideation behind the project is well documented at the repository.
Briefly, this project explores whether LLMs can better adopt collectives' perspectives by being fine-tuned on QA data generateed based on first-hand data belonging to the collectives of interest.
This model constitutes a prototype for this idea producing encouraging results.
