BEPA-7B-S2 is a GUI agent model fine-tuned from
UI-TARS-1.5-7B using the BEPA (Bi-Level Expert-to-Policy Assimilation) framework. This model achieves state-of-the-art performance among open-source end-to-end models on the OSWorld benchmark.
BEPA addresses two key challenges when using expert trajectories for training end-to-end GUI policies:
Transforms alien expert traces into policy-compatible trajectories by abstracting expert trajectories into compact natural-language plans, then letting the base policy act in the environment with plan conditioning.
Dynamically maintains a per-task cache, injecting guided trajectories into GRPO updates only upon total on-policy failure. The cache is continuously refreshed with the policy's own successful executions.
1@misc{wang2026offpolicyonpolicyenhancinggui,
2 title={From Off-Policy to On-Policy: Enhancing GUI Agents via Bi-level Expert-to-Policy Assimilation},
3 author={Zezhou Wang and Ziyun Zhang and Xiaoyi Zhang and Zhuzhong Qian and Yan Lu},
4 year={2026},
5 eprint={2601.05787},
6 archivePrefix={arXiv},
7 primaryClass={cs.AI},
8 url={https://arxiv.org/abs/2601.05787},
9}
This model is released under the
MIT License.
