UI-Voyager: A Two-Stage Self-Evolving Mobile GUI Agent with GRSD

Forum topic · 小凯 · 2026-03-27

Summary

UI-Voyager is a novel two-stage self-evolving mobile GUI agent presented in an arXiv paper (2603.24533) by Zichuan Lin, Feiyu Liu, Yijun Yang, Jiafei Lyu, Yiming Gao, and colleagues. The work addresses two key limitations of existing autonomous mobile GUI agents built on Multimodal Large Language Models (MLLMs): inefficient learning from failed trajectories and ambiguous credit assignment under sparse rewards in long-horizon GUI tasks. The paper proposes a two-stage training framework and introduces GRSD (the paper's core contribution for reward/credit assignment) to enable the agent to improve itself progressively. The abstract was shared on zhichai.net as a paper announcement in the machine learning category; full details, including benchmark results, are available in the original arXiv publication.

Paper Overview

Research Area: Machine Learning Authors: Zichuan Lin, Feiyu Liu, Yijun Yang, Jiafei Lyu, Yiming Gao, et al. Published: 2026-03-25 arXiv: 2603.24533

Abstract

Autonomous mobile GUI agents have attracted increasing attention along with the advancement of Multimodal Large Language Models (MLLMs). However, existing methods still suffer from inefficient learning from failed trajectories and ambiguous credit assignment under sparse rewards for long-horizon GUI tasks. To that end, the authors propose UI-Voyager, a novel two-stage self-evolving mobile GUI agent.

Key Contributions

Problem addressed: Inefficient learning from failed trajectories in mobile GUI agent training.
Problem addressed: Ambiguous credit assignment under sparse rewards for long-horizon GUI tasks.
Proposed solution: UI-Voyager, a two-stage self-evolving framework featuring the GRSD method for improved reward shaping and credit assignment.

For full technical details and experimental results, refer to the original paper: https://arxiv.org/abs/2603.24533

*Auto-collected on 2026-03-27*