Yue Wang

Quyu Kong, Xu Zhang, Zhenyu Yang et al.

Among existing online mobile-use benchmarks, AndroidWorld has emerged as the dominant benchmark due to its reproducible environment and deterministic evaluation; however, recent agents achieving over 90% success rates indicate its saturation and motivate the need for a more challenging benchmark. In addition, its environment lacks key application categories, such as e-commerce and enterprise communication, and does not reflect realistic mobile-use scenarios characterized by vague user instructions and hybrid tool usage. To bridge this gap, we introduce MobileWorld, a substantially more challenging benchmark designed to better reflect real-world mobile usage, comprising 201 tasks across 20 applications, while maintaining the same level of reproducible evaluation as AndroidWorld. The difficulty of MobileWorld is twofold. First, it emphasizes long-horizon tasks with cross-application interactions: MobileWorld requires nearly twice as many task-completion steps on average (27.8 vs. 14.3) and includes far more multi-application tasks (62.2% vs. 9.5%) compared to AndroidWorld. Second, MobileWorld extends beyond standard GUI manipulation by introducing novel task categories, including agent-user interaction and MCP-augmented tasks. To ensure robust evaluation, we provide snapshot-based container environment and precise functional verifications, including backend database inspection and task callback APIs. We further develop a planner-executor agentic framework with extended action spaces to support user interactions and MCP calls. Our results reveal a sharp performance drop compared to AndroidWorld, with the best agentic framework and end-to-end model achieving 51.7% and 20.9% success rates, respectively. Our analysis shows that current models struggle significantly with user interaction and MCP calls, offering a strategic roadmap toward more robust, next-generation mobile intelligence.

Debamita Ghosh, George K. Atia, Yue Wang

The deployment of reinforcement learning (RL) agents in real-world applications is often hindered by performance degradation caused by mismatches between training and deployment environments. Distributionally robust RL (DR-RL) addresses this issue by optimizing worst-case performance over an uncertainty set of transition dynamics. However, existing work typically relies on substantial prior knowledge-such as access to a generative model or a large offline dataset-and largely focuses on tabular methods that do not scale to complex domains. We overcome these limitations by proposing an online DR-RL algorithm with general function approximation that learns an optimal robust policy purely through interaction with the environment, without requiring prior models or offline data, enabling deployment in high-dimensional tasks. We further provide a theoretical analysis establishing a near-optimal sublinear regret bound under a total variation uncertainty set, demonstrating the sample efficiency and effectiveness of our method.