Curiosity Driven Knowledge Retrieval for Mobile Agents

2026-01-27

AI Summary

Mobile agents often underperform on complex, multi-step, and cross-application tasks due to incomplete knowledge and limited generalization capabilities. This work proposes a curiosity-driven dynamic knowledge retrieval mechanism that quantifies execution-time uncertainty to trigger the retrieval of external knowledge from documentation, codebases, and historical trajectories. The retrieved information is structured into AppCards to enhance the agent’s reasoning capabilities. By dynamically addressing knowledge gaps during task execution, the approach significantly improves performance. Evaluated on the AndroidWorld benchmark in conjunction with GPT-5, the method achieves an 88.8% success rate—outperforming existing approaches by an average of 6 percentage points and establishing a new state-of-the-art result.

Technology Category

• Multiagent Systems: Applications
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Mobile agents have made progress toward reliable smartphone automation, yet performance in complex applications remains limited by incomplete knowledge and weak generalization to unseen environments. We introduce a curiosity driven knowledge retrieval framework that formalizes uncertainty during execution as a curiosity score. When this score exceeds a threshold, the system retrieves external information from documentation, code repositories, and historical trajectories. Retrieved content is organized into structured AppCards, which encode functional semantics, parameter conventions, interface mappings, and interaction patterns. During execution, an enhanced agent selectively integrates relevant AppCards into its reasoning process, thereby compensating for knowledge blind spots and improving planning reliability. Evaluation on the AndroidWorld benchmark shows consistent improvements across backbones, with an average gain of six percentage points and a new state of the art success rate of 88.8\% when combined with GPT-5. Analysis indicates that AppCards are particularly effective for multi step and cross application tasks, while improvements depend on the backbone model. Case studies further confirm that AppCards reduce ambiguity, shorten exploration, and support stable execution trajectories. Task trajectories are publicly available at https://lisalsj.github.io/Droidrun-appcard/.

Problem

• mobile agents
• knowledge gaps
• generalization
• task planning
• smartphone automation

Innovation

• curiosity-driven retrieval
• AppCards
• mobile agents
• knowledge augmentation
• uncertainty-aware execution