Jaime Cernuda
Back to Publications
eScience'262026Conference

Agent Error Simulator: Fault Injection for Testing Recovery in Agentic Workflows

Keith Bateman, Jaime Cernuda, Luke Logan, Bogdan Nicolae, Franck Cappello, Xian-He Sun, Anthony Kougkas
Agentic WorkflowsFault InjectionBenchmarkingAI

Cite this work

@inproceedings{bateman2026aes,
  title={Agent Error Simulator: Fault Injection for Testing Recovery in Agentic Workflows},
  author={Bateman, Keith and Cernuda, Jaime and Logan, Luke and Nicolae, Bogdan and Cappello, Franck and Sun, Xian-He and Kougkas, Anthony},
  booktitle={22nd IEEE International Conference on eScience (eScience 2026)},
  year={2026}
}

Agent Error Simulator is a controllable fault-injection benchmark for agent-induced errors in multi-agent workflows, built around an iterative tool-calling computation that scales across agents. It injects logic, formatting, and tool-call errors at chosen times and roles so that recovery mechanisms can be measured despite stochastic agent behavior.