Wildertrek/personality-atlas-3072

Personality Atlas — 3072-dim Embeddings and Classifiers

Upgraded embeddings and retrained Random Forest classifiers for the 44-model personality atlas described in:

Raetano, J., Gregor, J., & Tamang, S. (2026). A Survey and Computational Atlas of Personality Models. ACM Transactions on Intelligent Systems and Technology (TIST). Under review.

Contents

Directory	Files	Description	Size
Embeddings_3072/	44 CSVs	OpenAI text-embedding-3-large (3072-dim) embeddings for all 6,694 trait rows	~440 MB
models_3072/	88 pkl files	Retrained scikit-learn RF classifiers + label encoders for all 44 models	~107 MB

Usage

from huggingface_hub import hf_hub_download
import pandas as pd
import joblib
import ast
import numpy as np

# Download a single model's assets
slug = "ocean"
emb_path = hf_hub_download("Wildertrek/personality-atlas-3072", f"Embeddings_3072/{slug}_embeddings.csv", repo_type="dataset")
model_path = hf_hub_download("Wildertrek/personality-atlas-3072", f"models_3072/{slug}_rf_model.pkl", repo_type="dataset")
enc_path = hf_hub_download("Wildertrek/personality-atlas-3072", f"models_3072/{slug}_label_encoder.pkl", repo_type="dataset")

emb_df = pd.read_csv(emb_path)
X = np.array([ast.literal_eval(e) for e in emb_df["Embedding"]])
model = joblib.load(model_path)
encoder = joblib.load(enc_path)

predictions = encoder.inverse_transform(model.predict(X))
print(f"{slug.upper()}: {len(X)} traits, accuracy = {(predictions == pd.read_csv(f'atlas/datasets/{slug}.csv')['Factor'].values).mean():.1%}")

Relationship to Main Repository

The 1536-dim assets (default) are in the main GitHub repository: github.com/Wildertrek/survey

The 3072-dim assets here are the upgraded versions from Experiment 2 (RQ7: embedding dimensionality upgrade). They improve mean accuracy from 58.7% to 63.8% (+5.1pp) across all 44 models.

Validation

# Clone the main repo and validate 3072-dim assets
git clone https://github.com/Wildertrek/survey.git
cd survey
python scripts/validate.py --all --dim 3072

License

MIT — same as the main repository.