Ella: Embodied Social Agents with Lifelong Memory

This repo contains codes for the following paper:

Hongxin Zhang*, Zheyuan Zhang*, Zeyuan Wang*, Zunzhe Zhang, Lixing Fang, Qinhong Zhou, Chuang Gan: Ella: Embodied Lifelong Learning Agents with Non-Parametric Memory

Paper: Arxiv

Project Website: Ella

Installation

Follow Virtual Community documents to install the environments our agents will live.

conda env create -f env.yaml
cd vico/Genesis
pip install -e .[dev]
cd agents/sg
./setup.sh

Run Experiments

The main implementation code of our Ella is in agents/ella.py.

We also prepare example scripts to run experiments under the folder scripts.

For example, to run experiments with Ella for one day in New York City,

./scripts/ODM/run_ella_odm_newyork.sh

To test Ella with Influence Battle Final,

./scripts/IB/test_IB_ella_newyork.sh

Citation

If you find our work useful, please consider citing:

@article{zhang2025ella,
  title={Ella: Embodied Social Agents with Lifelong Memory},
  author={Zhang, Hongxin and Zhang, Zheyuan and Wang, Zeyuan and Zhang, Zunzhe and Fang, Lixing and Zhou, Qinhong and Gan, Chuang},
  journal={arXiv preprint arXiv:2506.24019},
  year={2025}
}

Acknowledgement

We would like to acknowledge the Generative Agents codebase (https://github.com/joonspk-research/generative_agents) by Joon Sung Park and collaborators, which provided a foundational framework that greatly supported our work.