Research map / Vision-language and multimodal / Robot action and navigation
GoClick: Lightweight Element Grounding Model for Autonomous GUI Interaction
Hongxin Li, Yuntao Chen, Zhaoxiang Zhang
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 |
|---|---|---|---|
| apply coarse filtering to fine-tuning data (10.8M to 6.4M samples) (ablation) | GoClick-L · Seven grounding benchmarks (average) | accuracy: 71.6 -> 74.0 | Improved |
| apply fine filtering on top of coarse filtering (6.4M to 3.8M) (ablation) | GoClick-L · Seven grounding benchmarks (average) | accuracy: 74.0 -> 75.6 | Improved |
| remove outdated GUI pattern samples (ablation) | GoClick-L · MOTIF | accuracy: 74.0 -> 76.8 | Improved |
| remove functionality referring-expression generation samples (ablation) | GoClick-L · VWB EG | accuracy: 86.4 -> 90.2 | Improved |
| remove text REG samples (ablation) | GoClick-L · FuncPred | accuracy: 62.2 -> 62.2 | No clear effect |
| remove brief description REG samples (ablation) | GoClick-L · ScreenSpot-v2 | accuracy: 74.4 -> 74.0 | Hurt |
| use Florence-2 encoder-decoder instead of InternVL-2-1B decoder-only (ablation) | GoClick-L · FuncPred | accuracy: 21.9 -> 69.5 | Improved |
| use Florence-2 encoder-decoder (0.8B) instead of Qwen2VL-2B decoder-only (ablation) | GoClick-L · FuncPred | accuracy: 51.1 -> 69.5 | Improved |
| use Florence-2 0.8B instead of Qwen2VL-7B decoder-only (ablation) | GoClick-L · FuncPred | accuracy: 56.1 -> 69.5 | Improved |
| use Florence-2 0.8B instead of Qwen2VL-7B decoder-only (ablation) | GoClick-L · ScreenSpot | accuracy: 82.5 -> 78.5 | Hurt |
| scale GoClick from base (0.2B) to large (0.8B) | GoClick-L · FuncPred | accuracy: 64.4 -> 69.5 | Improved |
| add GoClick grounding to Gemini-2-Flash-Exp planner | Gemini-2-Flash-Exp planner · AITW overall | Step SR: 29.5 -> 47.2 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- UGround-7B — UGround: Towards Unified Visual Grounding with Unrolled Transformers (ICML 2026) · better on 2 of 2 shared results
- ShowUI-2B — ShowUI: One Vision-Language-Action Model for GUI Visual Agent (CVPR 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- GUI-Actor: Coordinate-Free Visual Grounding for GUI Agents (NeurIPS 2025 · similarity 0.91)
- Navigating the Digital World as Humans Do: Universal Visual Grounding for GUI Agents (ICLR 2025 · similarity 0.90)
- OS-ATLAS: Foundation Action Model for Generalist GUI Agents (ICLR 2025 · similarity 0.90)
- FocusUI: Efficient UI Grounding via Position-Preserving Visual Token Selection (CVPR 2026 · similarity 0.90)
- Scaling Computer-Use Grounding via User Interface Decomposition and Synthesis (NeurIPS 2025 · similarity 0.90)
- GUI-Spotlight: Adaptive Iterative Focus Refinement for Enhanced GUI Visual Grounding (ICML 2026 · similarity 0.89)
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.
