🧭 Lost in Recovery: Demystifying Recovery Behavior of Computer-Use Agents via Goal-Oriented Action Replay
A Scalable Framework for Systematic analysis of recovery behavior in computer-use agents via goal-oriented action replay
ReTRACE systematically analyzes the recovery behavior of computer-use agents by introducing Transition-Consistent Recovery and goal-oriented action replay, enabling reliable characterization of agents' ability to restore previous states after failures.
🚧 Code will be released soon.
If you find our research useful, please cite our paper.
@misc{bang2026lostinrecovery,
title={Lost in Recovery: Demystifying Recovery Behavior of Computer-Use Agents via Goal-Oriented Action Replay},
author={Byeonghun Bang and Sangwoo Park and Jinheon Baek and Kangsan Kim and Youngwan Lee and Sung Ju Hwang},
year={2026},
eprint={2610.27954},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2610.27954},
}