Executable Agentic Memory for GUI Agent

Modern GUI agents typically rely on a model-centric and step-wise interaction paradigm, where LLMs must re-interpret the UI and re-decide actions at every screen, which is fragile in long-horizon tasks. In this paper, we propose Executable Agentic Memory (EAM), a structured Knowledge Graph (KG) that shifts GUI planning from free-form generation to a robust retrieval-and-execution process. Our approach includes a sample-efficient memory construction pipeline using state-aware DFS and action-group mining to compress multi-step routines. To ensure efficient planning, we introduce a value-guided graph search where a lightweight Q-function model steers Monte Carlo Tree Search (MCTS) over the KG. We theoretically establish bias-consistency for the Q-model and derive sample complexity bounds for path recovery. Empirically, EAM outperforms state-of-the-art baselines like UI-TARS-7B by up to $19.6\%$ on AndroidWorld, while reducing token costs $6\times$ relative to GPT-4o. With a $2.8$s average latency, EAM enables reliable, quick, and long-horizon GUI automation.

AI agents are increasingly able to operate phones and websites, but long tasks are still difficult for them. A common approach is to look at each screen, ask a language model what to do next, and repeat this process step by step. This can be slow, costly, and unreliable, because one wrong interpretation can send the agent down the wrong path. We propose Executable Agentic Memory, a way for an agent to remember an app as a map of screens, actions, and valid transitions between them. Instead of guessing each next step from scratch, the agent searches this map to find an executable route to the user’s goal. It also learns which routes are more promising and groups common multi-step routines into reusable shortcuts. This makes GUI automation more reliable because the agent follows valid paths, and more efficient because it needs fewer expensive language-model calls. In our experiments, this led to faster execution, lower cost, and better task completion on standard mobile and web app benchmarks.