UI-Oceanus: Scaling GUI Agents with Synthetic Environmental Dynamics
Mengzhou Wu , Yuzhe Guo , Yuan Cao , Haochuan Lu , Songhe Zhu , Pingzhe Qu , Xin Chen , Kang Qin , Zhongpu Wang , Xiaode Zhang , Xinyi Wang , Wei Dai , Gang Cao , Yuetang Deng , Zhi Gong , Dezhi Ran , Linyi Li , Wei Yang , Tao Xie
- 🏛 Institutions
- PKU , Tencent
- 📅 Date
- February 11, 2026
- 📑 Publisher
- arXiv
- 💻 Env
- Mobile
- 🔑 Keywords
TLDR
UI-Oceanus shifts learning from mimicking trajectories to mastering interaction physics via forward dynamics prediction. Using synthetic environmental dynamics as supervision, it improves GUI agents by 7% on offline benchmarks and 16.8% in online navigation, with performance scaling log-linearly with data volume.
Related papers (24)
- AutoGUIWorld: Image Generators as Visual World Models for GUI AgentOctober 1, 2026 · arXiv
- WebWorld: A Large-Scale World Model for Web Agent TrainingFebruary 16, 2026 · arXiv
- Beyond Next-Screen Prediction: The Decision Value of GUI World ModelsOctober 17, 2026 · arXiv
- Right Screen, Wrong Transition: World Models as Verifiers for GUI AgentsOctober 8, 2026 · arXiv
- Imagine to Act: High-Fidelity Data Synthesis via Image Editing World Model for Scalable GUI Agent TrainingOctober 5, 2026 · arXiv
- Learn How to Act from Your Own Interactions: On-Policy Self-Distillation for GUI AgentsSeptember 23, 2026 · arXiv
- Towards Omni-dimensional GUI Agent Navigation with Masked Trajectory PredictionSeptember 22, 2026 · EMNLP 2026
- MATE: Policy-Aware Security Auditing for Mobile Agents via Synthesis-Driven Trajectory LearningSeptember 19, 2026 · USENIX Security 2026
- MintAct: A Unified Visual Agent for Digital EnvironmentsSeptember 18, 2026 · arXiv
- BlueLM-GUI Technical Report: A Real-Device-Centric Flywheel for Self-Improving Mobile GUI AgentsSeptember 11, 2026 · arXiv
- LLaDA-UI: Bringing Block-wise Diffusion to Vision-Language GUI AgentsSeptember 9, 2026 · arXiv
- GUI-CC: Benchmarking Contextual Consistency of GUI World Models as Agent EnvironmentsAugust 30, 2026 · arXiv
- WM-R1: Training GUI Agents to Reason and leverage World Models with Reinforcement LearningAugust 27, 2026 · arXiv
- AppDeltaWorld: Transition-Grounded Delta Code World Model for Mobile GUI AgentsAugust 6, 2026 · arXiv
- Qwen-UI-Agent Technical Report: Toward Next-Generation Real-World Centric Foundation GUI AgentsJuly 30, 2026 · arXiv
- Scaling GUI Agents with Visual State TransitionsJuly 27, 2026 · arXiv
- SEE: Structure-aware Exploring and Exploiting for Long-horizon GUI Agent Trajectory SynthesisJuly 20, 2026 · arXiv
- SeerGuard: A Safety Framework for Mobile GUI Agents via World Model PredictionJuly 17, 2026 · arXiv
- Xiaomi-GUI-0 Technical ReportJune 30, 2026 · arXiv
- How Mobile World Model Guides GUI Agents?May 11, 2026 · arXiv
- ClawGUI: A Unified Framework for Training, Evaluating, and Deploying GUI AgentsApril 13, 2026 · arXiv
- SecAgent: Efficient Mobile GUI Agent with Semantic ContextMarch 9, 2026 · arXiv
- CoME: Empowering Channel-of-Mobile-Experts with Informative Hybrid-Capabilities ReasoningFebruary 27, 2026 · ICML 2026 (Poster)
- Mobile-Agent-v3.5: Multi-platform Fundamental GUI AgentsFebruary 15, 2026 · arXiv