Skip to content

fouratifares/coherence-agi

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Coherence-based measure of Artificial General Intelligence (AGI)

arXiv Medium

Implementation of the coherence-based AGI measure introduced in
Fourati (2025), A Coherence-Based Measure of AGI.


🔍 Overview

The Coherent-AGI library provides two evaluation metrics:

1️⃣ AGIₚ — Generalized (Power) Mean

A parametric mean controlling compensability across cognitive domains.

2️⃣ AGI_AUC — AUC Across Compensability Exponents

Measures coherence by integrating AGIₚ across coupling regimes.

This metric penalizes structural imbalance and brittleness across domains.

Why coherence?

Arithmetic aggregation overstates general intelligence, because strong domains can fully compensate for collapsed ones.
AGIₚ and AGI_AUC fix this by penalizing uneven ability profiles.


📦 Installation

Local install:

pip install -e .

🚀 Usage

from coherent_agi import agi_p, agi_auc

# Domain scores in [0,100]:
scores = [90, 100, 100, 70, 50, 0, 40, 40, 60, 30]

print("AGI_p (p=1): ", agi_p(scores, 1))     # arithmetic mean
print("AGI_p (p=0): ", agi_p(scores, 0))     # geometric mean
print("AGI_p (p=-1):", agi_p(scores, -1))    # harmonic mean

# Full coherence-based AGI:
print("AGI_AUC:      ", agi_auc(scores))

📚 Citation

If you use this library or the AGI metrics in a publication, please cite:

@article{fourati2025coherence,
  title={A Coherence-Based Measure of AGI},
  author={Fourati, Fares},
  journal={arXiv preprint arXiv:2510.20784},
  year={2025}
}

About

A Coherence-Based Measure of AGI

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages