PreView: Training-Free Preprocessor Boosts GUI Grounding via Zoomed Views and Cues
Wenjun Hou · hf · 2026-10-30
- Independent researchers Wenjun Hou and Kaishuai Xu released the paper PreView: Enhancing GUI Grounding with Preprocessed Views and Cues (arXiv, Oct 24, 2026).
- GUI grounding requires locating small controls in cluttered screenshots; cropping/zooming exposes detail but often leaves ambiguity about the intended target.
- PreView is an auxiliary preprocessor that leaves the frozen grounding model untouched and improves its inputs instead, adding two complementary features: task-relevant views (cropping out irrelevant content) and visual-textual cues clarifying the target control.
- The "modify inputs, not the model" approach boosts grounding accuracy without retraining — practically relevant for GUI agent pipelines.
More from coding & agent
- VibeGame from Nanjing University builds complete 2D games from prompts with adversarial agent teams — nanjinguniv · 2026-10-30
- WebAnything Auto-Rebuilds Real Websites into Scalable Web Agent Environments — Tianci Xie · 2026-10-30
- WorkFactory Scales Professional Task Generation for Agent Training, Yields WorkPilot — IQuestLab · 2026-10-30
- Kimi launches 3D Make: connects to Blender and CAD tools to build models, scenes, and games — Kimi_Moonshot · 2026-10-30
- Animated short made with ComfyUI: ChatGPT poses, Qwen restyle, MiniMax H3 — bseoan · 2026-10-30
- Claude API docs reveal beta Agents, Sessions and Environments APIs with managed sandboxes — aigclink · 2026-10-30