LLMs as Scalable, General-Purpose Simulators For Evolving Digital Agent Training
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866912652951814144 |
|---|---|
| author | Wang, Yiming Yin, Da Cui, Yuedong Zheng, Ruichen Li, Zhiqian Lin, Zongyu Wu, Di Wu, Xueqing Ye, Chenchen Zhou, Yu Chang, Kai-Wei |
| author_facet | Wang, Yiming Yin, Da Cui, Yuedong Zheng, Ruichen Li, Zhiqian Lin, Zongyu Wu, Di Wu, Xueqing Ye, Chenchen Zhou, Yu Chang, Kai-Wei |
| contents | Digital agents require diverse, large-scale UI trajectories to generalize across real-world tasks, yet collecting such data is prohibitively expensive in both human annotation, infra and engineering perspectives. To this end, we introduce $\textbf{UI-Simulator}$, a scalable paradigm that generates structured UI states and transitions to synthesize training trajectories at scale. Our paradigm integrates a digital world simulator for diverse UI states, a guided rollout process for coherent exploration, and a trajectory wrapper that produces high-quality and diverse trajectories for agent training. We further propose $\textbf{UI-Simulator-Grow}$, a targeted scaling strategy that enables more rapid and data-efficient scaling by prioritizing high-impact tasks and synthesizes informative trajectory variants. Experiments on WebArena and AndroidWorld show that UI-Simulator rivals or surpasses open-source agents trained on real UIs with significantly better robustness, despite using weaker teacher models. Moreover, UI-Simulator-Grow matches the performance of Llama-3-70B-Instruct using only Llama-3-8B-Instruct as the base model, highlighting the potential of targeted synthesis scaling paradigm to continuously and efficiently enhance the digital agents. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_14969 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | LLMs as Scalable, General-Purpose Simulators For Evolving Digital Agent Training Wang, Yiming Yin, Da Cui, Yuedong Zheng, Ruichen Li, Zhiqian Lin, Zongyu Wu, Di Wu, Xueqing Ye, Chenchen Zhou, Yu Chang, Kai-Wei Computation and Language Artificial Intelligence Machine Learning Digital agents require diverse, large-scale UI trajectories to generalize across real-world tasks, yet collecting such data is prohibitively expensive in both human annotation, infra and engineering perspectives. To this end, we introduce $\textbf{UI-Simulator}$, a scalable paradigm that generates structured UI states and transitions to synthesize training trajectories at scale. Our paradigm integrates a digital world simulator for diverse UI states, a guided rollout process for coherent exploration, and a trajectory wrapper that produces high-quality and diverse trajectories for agent training. We further propose $\textbf{UI-Simulator-Grow}$, a targeted scaling strategy that enables more rapid and data-efficient scaling by prioritizing high-impact tasks and synthesizes informative trajectory variants. Experiments on WebArena and AndroidWorld show that UI-Simulator rivals or surpasses open-source agents trained on real UIs with significantly better robustness, despite using weaker teacher models. Moreover, UI-Simulator-Grow matches the performance of Llama-3-70B-Instruct using only Llama-3-8B-Instruct as the base model, highlighting the potential of targeted synthesis scaling paradigm to continuously and efficiently enhance the digital agents. |
| title | LLMs as Scalable, General-Purpose Simulators For Evolving Digital Agent Training |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2510.14969 |