Kuumba Regressor 🖌️
This model predicts a Kuumba (Creativity) score for text passages, where Kuumba is defined as:
"Visionary inventor and artist, bringing innovation, beauty, and cultural tools for transformation."
The model outputs a continuous score between 0.0 – 1.0, with higher values representing stronger expressions of Kuumba (creativity, innovation, and transformative impact).
Model Details
- Developed by: Francky Clerge (ClergeF) as part of the MyVillage Soulprint Project
- Model type: RoBERTa-based regression model
- Language(s): English
- License: MIT (recommended, adjust if needed)
- Finetuned from model:
roberta-base
Model Sources
- Repository: ClergeF/kuumba-regressor
Uses
Direct Use
- Estimate a Kuumba creativity score for any input text.
- Intended for research, education, and experimentation in cultural/creative AI.
Downstream Use
- Integration into the Soulprint framework to measure community values.
- Can be extended to other Soulprint dimensions (Ubuntu, Sankofa, etc.).
Out-of-Scope Use
- Not intended as a definitive measurement of human creativity.
- Should not be used in decision-making contexts (e.g., hiring, grading, funding).
Bias, Risks, and Limitations
- Training data was mock-generated for testing. Scores reflect synthetic judgments, not universal truths.
- Creativity is subjective and culturally contextual.
- Model may overfit to phrasing patterns (e.g., technical vs artistic descriptions).
Recommendation: Use this as an exploratory tool, not an authoritative metric.
How to Get Started
Load the model in Python:
from transformers import AutoTokenizer, AutoModelForSequenceClassification
model_name = "ClergeF/kuumba-regressor"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)
inputs = tokenizer("We built an augmented reality mural that reveals poems when scanned.", return_tensors="pt")
with torch.no_grad():
outputs = model(**inputs)
score = outputs.logits.squeeze().item()
print("Predicted Kuumba score:", score)
- Downloads last month
- 11