Search papers, labs, and topics across Lattice.
This paper introduces VeriGUI, a Verification-driven GUI Agent that explicitly models action outcomes and recovery in noisy GUI environments. VeriGUI employs a Thinking-Verification-Action-Expectation (TVAE) framework to detect action failures and guide corrective reasoning. A two-stage training pipeline, combining Robust SFT with synthetic failure trajectories and GRPO with asymmetric verification rewards, enhances robustness.
Autonomous GUI agents can now handle noisy real-world conditions, like network latency and system interruptions, thanks to a new framework that verifies action success and self-corrects.
Autonomous GUI agents based on vision-language models (VLMs) often assume deterministic environment responses, generating actions without verifying whether previous operations succeeded. In real-world settings with network latency, rendering delays, and system interruptions, this assumption leads to undetected action failures, repetitive ineffective behaviors, and catastrophic error accumulation. Moreover, learning robust recovery strategies is challenging due to the high cost of online interaction and the lack of real-time feedback in offline datasets.We propose VeriGUI (Verification-driven GUI Agent), which explicitly models action outcomes and recovery under noisy environments. VeriGUI introduces a Thinking--Verification--Action--Expectation (TVAE) framework to detect failures and guide corrective reasoning, and a two-stage training pipeline that combines Robust SFT with synthetic failure trajectories and GRPO with asymmetric verification rewards. We further construct a Robustness Benchmark based on AndroidControl to evaluate failure recognition and correction. Experiments show that VeriGUI significantly reduces failure loops and improves recovery success while maintaining competitive standard task performance.
