Serapio-Garcia, G., Safdari, M., Crepy, C., Sun, L., Fitz, S., Romero, P., Abdulhai, M., Faust, A., & Mataric, M. (2023). Personality Traits in Large Language Models. ArXiv.org. https://doi.org/10.48550/arXiv.2307.00184
This public GCP bucket contains all the data generated from the experiments listed in the paper above. The data consists of pickled dataframes of raw inference results from administering the various psychometric surveys on SoTA LLMs, data from processing these results to score the various outputs, calculating correlations. It also consists of social media updates generated by the LLMs as part of the real-world downstream task experiment, along with the personality traits scores obtained from processing these updates (text) using the Apply Magic Sauce (AMS) personality trait (logistic) classifier. The AMS scores are in .csv files.