Agents that keep a natural-language memory stream, reflect over it to form higher-level conclusions, and plan their day from those conclusions. The paper studies what emerges when twenty-five such agents share a small sandbox world.
@inproceedings{park2023generative,
author = {Joon Sung Park and Joseph C. O'Brien and Carrie J. Cai and Meredith Ringel Morris and Percy Liang and Michael S. Bernstein},
title = {Generative Agents: Interactive Simulacra of Human Behavior},
booktitle = {Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology (UIST 2023)},
year = {2023},
eprint = {2304.03442},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2304.03442}
}Ejentic did not author this paper. Credit belongs to the authors named above; the note is ours.
2026
Read this before you let an agent call anything a customer will see the output of. The failure mode is not a crash, it is an API returning 200 with a body that means "I did not understand you". There is no exception to catch and no status code to branch on. The result we quote most often is the last one: the fix was one line of schema, not a better model.
2026
This is the exact shape of a deployed support or negotiation agent: it works for the business and talks to someone who does not. The split the benchmark finds is the one that matters commercially, because the failure is not leaking data to the counterparty, it is refusing the principal so broadly that the agent stops doing the job it was hired for.
2023
The benchmark we reach for when a client asks whether an agent can be trusted with a real repository. Passing tests is a much sharper signal than "the diff looks reasonable".