Research map / Multi-agent systems and games / LLM agents
Less is More: Empowering GUI Agent with Context-Aware Simplification
Gongwei Chen, Xurui Zhou, Rui Shao, Yibo Lyu, Kaiwen Zhou, Shuai Wang, Wentao Li, Yinchuan Li, Zhongang Qi, Liqiang Nie
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add history compression by directly dropping history vision tokens (ablation) | Qwen2VL-2B · AITW | step SR: 69.0 -> 67.3 | Hurt |
| add history compression by directly dropping history vision tokens (ablation) | Qwen2VL-2B · GUI-Odyssey | step SR: 74.9 -> 71.8 | Hurt |
| add consistency-guided training to history compression (ablation) | Qwen2VL-2B · AITW | step SR: 67.3 -> 68.9 | Improved |
| add consistency-guided training to history compression (ablation) | Qwen2VL-2B · GUI-Odyssey | step SR: 71.8 -> 73.7 | Improved |
| add masking-based element pruning (ablation) | SimpAgent · AITW | step SR: 68.9 -> 71.3 | Improved |
| add masking-based element pruning (ablation) | SimpAgent · GUI-Odyssey | step SR: 73.7 -> 76.0 | Improved |
| add 4 previous actions to current observation | Qwen2-VL-2B · AITW | step SR: 62.5 -> 66.0 | Improved |
| add 4 previous observations to action history | Qwen2-VL-2B · AITW | step SR: 66.0 -> 69.0 | Improved |
| keep only the half of screenshot containing the click point | Qwen2-VL-2B · AITW | step SR: 66.0 -> 68.8 | Improved |
| drop history tokens after layer 12 instead of layer 3 (ablation) | Qwen2VL-2B (no consistency) · AITW | step SR: 67.3 -> 67.4 | No clear effect |
| replace FastV-0 compression with consistency-guided compression | Qwen2VL-2B · AITW | step SR: 63.8 -> 68.9 | Improved |
| use Victor summary tokens instead of consistency-guided compression | Qwen2VL-2B · AITW | step SR: 68.9 -> 67.6 | Hurt |
Closest papers by meaning
- HiconAgent: History Context-aware Policy Optimization for GUI Agents (CVPR 2026 · similarity 0.89)
- ShowUI: One Vision-Language-Action Model for GUI Visual Agent (CVPR 2025 · similarity 0.87)
- GUI-Xplore: Empowering Generalizable GUI Agents with One Exploration (CVPR 2025 · similarity 0.86)
- History-Aware Reasoning for GUI Agents (AAAI 2026 · similarity 0.86)
- GUI Exploration Lab: Enhancing Screen Navigation in Agents via Multi-Turn Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- GUI-Rise: Structured Reasoning and History Summarization for GUI Navigation (NeurIPS 2025 · similarity 0.85)
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.
