AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

GUI-SAGE: Enhancing GUI Automation with Self-Explanatory Learning

Fei Tang, Zhangxuan Gu, Zhengxi Lu, Shangzhan Zhang, Zhengwen Zeng, Shuheng Shen, Changhua Meng, Yuchen Yan, Wenqi Zhang, Yongliang Shen, Weiming Lu, Yueting Zhuang

CVPR 2026 · Reinforcement learning › Imitation learning · approach: Imitation combined with RL

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add self-explanation samples conditioned on ground-truth action to GRPO rollouts (ablation)Vanilla-GRPO (Qwen2.5-VL-3B) · AndroidControl-Low + HighAvg SR (%): 80.1 -> 83.1Improved
add expert demonstrations from Qwen2.5-VL-72B instead of self-explanations (ablation)Vanilla-GRPO (Qwen2.5-VL-3B) · AndroidControl-Low + HighAvg SR (%): 80.1 -> 81.2Improved
replace expert demonstrations with self-explanations (ablation)GUI-SAGE-3B · AndroidControl-Low + HighAvg SR (%): 81.2 -> 83.1Improved
add Entropy-Modulated Credit Assignment (EMCA) to self-explanation (ablation)GUI-SAGE-3B · AndroidControl-Low + HighAvg SR (%): 83.1 -> 84.2Improved
add EMCA to vanilla GRPO (ablation)Vanilla-GRPO · AndroidControl-Low + HighAvg SR (%): 80.1 -> 80.6Improved
add EMCA to expert demonstration training (ablation)Expert-CoT GRPO · AndroidControl-Low + HighAvg SR (%): 81.2 -> 81.9Improved
add explicit reasoning before action prediction (ablation)Vanilla-GRPO · AndroidControl-Low + HighAvg SR (%): 78.5 -> 80.1Improved
add partial hint of action type only (ATH) (ablation)Vanilla-GRPO · AndroidControl-Low + HighAvg SR (%): 80.1 -> 81.4Improved
add partial hint of action parameters only (APH) (ablation)Vanilla-GRPO · AndroidControl-Low + HighAvg SR (%): 80.1 -> 81.8Improved
use complete ground-truth action hint instead of partial hints (ablation)hint-based GRPO · AndroidControl-Low + HighAvg SR (%): 81.8 -> 83.1Improved
set reward weights to equal w1=w2=1.0 instead of 0.8/0.2 or 0.2/0.8 (ablation)GUI-SAGE-3B · AndroidControl + GUI-OdysseyAvg SR (%): 83.5 / 83.7 -> 84.2No clear effect
measure zero-advantage fraction in vanilla GRPO early trainingVanilla-GRPO · AndroidControl/GUI-Odyssey training, steps 0-100zero-advantage samples (%): 73.2Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy