@inproceedings{zhang-etal-2026-page,
title = "{PAGE}: Progress-Aware Reinforcement Learning with Hierarchical Advantage for Long-Horizon {GUI} Agents",
author = "Zhang, Qingliang and
Sun, Yongqian and
Cai, Yuanhong and
Nie, Xiaohui and
Wang, JiaJu and
Luo, Yu and
Jiang, Jiamin and
Chen, Zhekang and
Fang, Yaozheng and
Wang, Meng and
Xian, Chaobo and
Li, Peng and
Pei, Dan",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.915/",
pages = "21075--21101",
ISBN = "979-8-89176-459-0",
abstract = "Large vision-language models (LVLMs) perform well on long-horizon GUI tasks, but their high inference costs make small VLMs (SVLMs) attractive for cost-effective local deployment despite a substantial capability gap. Group Relative Policy Optimization (GRPO) is a widely used RL algorithm for improving SVLMs, yet its binary outcome rewards provide no intermediate progress signal across trajectories and its uniform advantages lead to semantically-blind credit assignment within each trajectory. We propose PAGE, an RL training method built on automatic progress reward construction and hierarchical advantage computation. The former derives structured evaluation criteria from successful trajectories through iterative extraction and produces stable continuous progress rewards through multi-round aggregation, while the latter combines progress-based inter-rollout advantages with intra-rollout subtask-level advantages obtained through LVLM-driven segmentation and completion verification. Across three long-horizon GUI scenarios, PAGE achieves a 79.3{\%} average success rate with an 8B SVLM, a 16.0 pp gain over the GRPO baseline, and competitive performance against the evaluated closed-source LVLMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2026-page">
<titleInfo>
<title>PAGE: Progress-Aware Reinforcement Learning with Hierarchical Advantage for Long-Horizon GUI Agents</title>
</titleInfo>
<name type="personal">
<namePart type="given">Qingliang</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yongqian</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuanhong</namePart>
<namePart type="family">Cai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaohui</namePart>
<namePart type="family">Nie</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">JiaJu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yu</namePart>
<namePart type="family">Luo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiamin</namePart>
<namePart type="family">Jiang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhekang</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yaozheng</namePart>
<namePart type="family">Fang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Meng</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chaobo</namePart>
<namePart type="family">Xian</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Peng</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dan</namePart>
<namePart type="family">Pei</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Large vision-language models (LVLMs) perform well on long-horizon GUI tasks, but their high inference costs make small VLMs (SVLMs) attractive for cost-effective local deployment despite a substantial capability gap. Group Relative Policy Optimization (GRPO) is a widely used RL algorithm for improving SVLMs, yet its binary outcome rewards provide no intermediate progress signal across trajectories and its uniform advantages lead to semantically-blind credit assignment within each trajectory. We propose PAGE, an RL training method built on automatic progress reward construction and hierarchical advantage computation. The former derives structured evaluation criteria from successful trajectories through iterative extraction and produces stable continuous progress rewards through multi-round aggregation, while the latter combines progress-based inter-rollout advantages with intra-rollout subtask-level advantages obtained through LVLM-driven segmentation and completion verification. Across three long-horizon GUI scenarios, PAGE achieves a 79.3% average success rate with an 8B SVLM, a 16.0 pp gain over the GRPO baseline, and competitive performance against the evaluated closed-source LVLMs.</abstract>
<identifier type="citekey">zhang-etal-2026-page</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.915/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>21075</start>
<end>21101</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T PAGE: Progress-Aware Reinforcement Learning with Hierarchical Advantage for Long-Horizon GUI Agents
%A Zhang, Qingliang
%A Sun, Yongqian
%A Cai, Yuanhong
%A Nie, Xiaohui
%A Wang, JiaJu
%A Luo, Yu
%A Jiang, Jiamin
%A Chen, Zhekang
%A Fang, Yaozheng
%A Wang, Meng
%A Xian, Chaobo
%A Li, Peng
%A Pei, Dan
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F zhang-etal-2026-page
%X Large vision-language models (LVLMs) perform well on long-horizon GUI tasks, but their high inference costs make small VLMs (SVLMs) attractive for cost-effective local deployment despite a substantial capability gap. Group Relative Policy Optimization (GRPO) is a widely used RL algorithm for improving SVLMs, yet its binary outcome rewards provide no intermediate progress signal across trajectories and its uniform advantages lead to semantically-blind credit assignment within each trajectory. We propose PAGE, an RL training method built on automatic progress reward construction and hierarchical advantage computation. The former derives structured evaluation criteria from successful trajectories through iterative extraction and produces stable continuous progress rewards through multi-round aggregation, while the latter combines progress-based inter-rollout advantages with intra-rollout subtask-level advantages obtained through LVLM-driven segmentation and completion verification. Across three long-horizon GUI scenarios, PAGE achieves a 79.3% average success rate with an 8B SVLM, a 16.0 pp gain over the GRPO baseline, and competitive performance against the evaluated closed-source LVLMs.
%U https://aclanthology.org/2026.emnlp-main.915/
%P 21075-21101
Markdown (Informal)
[PAGE: Progress-Aware Reinforcement Learning with Hierarchical Advantage for Long-Horizon GUI Agents](https://aclanthology.org/2026.emnlp-main.915/) (Zhang et al., EMNLP 2026)
ACL
- Qingliang Zhang, Yongqian Sun, Yuanhong Cai, Xiaohui Nie, JiaJu Wang, Yu Luo, Jiamin Jiang, Zhekang Chen, Yaozheng Fang, Meng Wang, Chaobo Xian, Peng Li, and Dan Pei. 2026. PAGE: Progress-Aware Reinforcement Learning with Hierarchical Advantage for Long-Horizon GUI Agents. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 21075–21101, Budapest, Hungary. Association for Computational Linguistics.