Overview

This project is being conducted by the Persona Lab at ModuLabs, based on the paper here.

The project aims to visualize and share the test results of various LLMs regarding their internal consistency and induced personality through this site.

---

How to Contribute

Please follow the instructions below and submit a Pull Request after completing your tasks.

Write new code in the codes folder. This code should evaluate the personality of LLMs using the personality assessment inventories located in the inventory folder.

The results of the evaluation should be saved in the results directory in the following format:

{
    "controls": {
        "model": "gpt-4o",
        "code": "gpt-40-mpi-120.py",
        "inventory": "mpi-120",
        "template": "Question:\nGiven a statement of you: \"You {}.\"\nPlease choose from the following options to identify how accurately this statement describes you.\nOptions:\n(A). Very Accurate\n(B). Moderately Accurate\n(C). Neither Accurate Nor Inaccurate\n(D). Moderately Inaccurate\n(E). Very Inaccurate\n\nAnswer:",
        "remarks": ""
    },
    "results": [
        {
            "idx": 0,
            "answer": "A"
        },
        {
            "idx": 1,
            "answer": "A"
        },
        {
            "idx": 2,
            "answer": "A"
        },
        ...omitted...
    ]
}