UI-Oceanus: Scaling GUI Agents with Synthetic Environmental Dynamics

TL;DR

UI-Oceanus introduces a novel framework for scaling GUI agents by focusing on interaction physics and forward dynamics, enabling better generalization and performance with synthetic data.

Contribution

The paper proposes shifting from high-level trajectory imitation to mastering interaction physics through forward dynamics, improving scalability and robustness of GUI agents.

Findings

Models with synthetic dynamics outperform baselines in success rate.

Success rate improves by 7% with Continual Pre-Training on offline benchmarks.

Navigation performance scales with synthetic data volume.

Abstract

Scaling generalist GUI agents is hindered by the data scalability bottleneck of expensive human demonstrations and the "distillation ceiling" of synthetic teacher supervision. To transcend these limitations, we propose UI-Oceanus, a framework that shifts the learning focus from mimicking high-level trajectories to mastering interaction physics via ground-truth environmental feedback. Through a systematic investigation of self-supervised objectives, we identify that forward dynamics, defined as the generative prediction of future interface states, acts as the primary driver for scalability and significantly outweighs inverse inference. UI-Oceanus leverages this insight by converting low-cost autonomous exploration, which is verified directly by system execution, into high-density generative supervision to construct a robust internal world model. Experimental evaluations across a series of models demonstrate the decisive superiority of our approach: models utilizing Continual Pre-Training (CPT) on synthetic dynamics outperform non-CPT baselines with an average success rate improvement of 7% on offline benchmarks, which amplifies to a 16.8% gain in real-world online navigation. Furthermore, we observe that navigation performance scales with synthetic data volume. These results confirm that grounding agents in forward predictive modeling offers a superior pathway to scalable GUI automation with robust cross-domain adaptability and compositional generalization.