Faithful Mobile GUI Agents with Guided Advantage Estimator

TL;DR

This paper introduces Faithful-Agent, a framework for GUI interaction that enhances faithfulness and evidence grounding using a two-stage training process and a novel guided advantage estimator, significantly improving task success rates.

Contribution

It presents a new faithfulness-first approach with a guided advantage estimator to improve GUI agent reliability and grounding, outperforming previous methods.

Findings

Trap SR increased from 13.88% to 80.21% with Faithful-Agent.

Faithful-Agent maintains strong instruction-following performance.

GuAE prevents advantage collapse in low-variance rollout groups.

Abstract

Vision-language model based graphical user interface (GUI) agents have shown strong interaction capabilities. However, they often behave unfaithfully, relying on memorized shortcuts rather than grounding actions in displayed screen evidence or user instructions. To address this, we propose Faithful-Agent, a faithfulness-first framework that reformulates GUI interaction to prioritize evidence groundedness and internal consistency. Faithful-Agent employs a two-stage pipeline: (i) a faithfulness-oriented SFT stage to instill abstainment behaviors under evidence perturbations; (ii) an RFT stage that further amplifies faithfulness by introducing the guided advantage estimator (GuAE), an anchor-based and variance-adaptive advantage tempering mechanism built upon GRPO. GuAE prevents advantage collapse in low-variance rollout groups under sparse GUI rewards, and with a thought-action consistency reward, Faithful-Agent (Stage II) elevates the Trap SR from 13.88\% to 80.21\% relative to the baseline, while preserving robust general instruction-following performance.