Search papers, labs, and topics across Lattice.
This paper introduces AnTrap, a novel benchmark designed to systematically evaluate the robustness of Android GUI agents against dynamic runtime anomalies. By categorizing real-world anomalies into a four-layer taxonomy and injecting these perturbations into agent execution, the study reveals that even the most advanced GUI models exhibit significant vulnerabilities, particularly to deep contextual traps. The findings highlight that while some anomalies can be mitigated through adversarial reinforcement learning, intrinsic limitations persist that cannot be overcome merely by training in adversarial environments.
Even the strongest Android GUI agents are universally vulnerable to runtime anomalies, revealing critical flaws in their robustness.
GUI agents often encounter dynamic anomalies when deployed on Android devices, from unexpected pop-ups to action misuse, yet existing benchmarks lack systematic evaluation of agent robustness against runtime anomalies. We introduce AnTrap, a comprehensive benchmark that injects dynamic perturbations into agent execution trajectories. We propose a taxonomy organizing real-world anomalies into four layers (State, Thinking, Action and Round) with ten fine-grained subcategories, and develop a construction pipeline that preserves task solvability while introducing realistic adversarial conditions. Evaluating 16 leading GUI models, we reveal universal vulnerability to dynamic anomalies, with even the strongest models suffering significant performance degradation. Furthermore, we conduct GRPO training in both original and adversarial environments to validate our benchmark, separating environment-learnable anomalies from reasoning-bottlenecked ones. Our findings show that while single-step traps at state and action layers are largely addressable through adversarial reinforcement learning, deep contextual traps, like state deadlock, expose intrinsic limitations that cannot be resolved by training in environments with traps alone.
