Graph Meme Pool (GMP)

A Computational Framework for Simulating the Evolution of Textual Content on Graphs

About The Framework

Graph Meme Pool (GMP) is a computational framework for simulating the evolution of text by modeling textual units as memes — replicable cultural information analogous to genes. Memes occupy nodes in a graph where edges represent pathways of interaction. They propagate probabilistically along these edges, after which a fitness function evaluates them to guide a selection process. This selection determines whether Large Language Models (LLMs) introduce variation by merging memes, adapting content during transmission. Through iterative cycles of propagation, selection, and variation, the system models how memes evolve, spread, and adapt within networked populations that can themselves change over time.

Code Repositories

Main Repository (Latest Version)

The latest, actively developed version of the framework. Features dynamic graphs, advanced selection mechanisms, resumable simulations, and a robust experiment management system.

View on GitHub

Archival Repository (v1)

An archive containing the source code and configuration used for the original research paper, "Selection in the Meme Pool: Graph-Based Evolution of Textual Content".

View v1 on GitHub

Visualizations

Dynamic Simulation Replay

An interactive web-based visualization that replays a simulation generation by generation. Watch as meme content evolves within nodes and the network topology itself changes through node creation and removal.

Launch Visualization

Publications

Selection in the Meme Pool: Graph-Based Evolution of Textual Content (ECAI 2025)

The research paper that introduced the foundational concepts of the Graph Meme Pool framework.

View Paper (Link to follow)

Citation

If you use the Graph Meme Pool (GMP) framework in your research, please cite the original paper which introduced its foundational concepts:

@inproceedings{babic2025selection,
  author    = {Babi\'{c}, Karlo},
  title     = {Selection in the Meme Pool: Graph-Based Evolution of Textual Content},
  booktitle = {Proceedings of the 28th European Conference on Artificial Intelligence (ECAI 2025)},
  year      = {2025},
  address   = {Bologna, Italy},
  note      = {Accepted for publication}
}