Research map / Adversarial robustness and security / Attacks on text and graphs
LaSM: Layer-wise Scaling Mechanism for Defending Pop-up Attack on GUI Agents
Zihe Yan, Zhuosheng Zhang, Jiaping Gui, Gongshen Liu
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 attention scaling to the MLP scaling (scale both modules) (ablation) | LaSM on Qwen2-VL-7B · Pop-up screenshots | DSR: 0.47 -> 84.80 | Improved |
| add MLP scaling to the attention scaling (scale both modules) (ablation) | LaSM on Qwen2-VL-7B · Pop-up screenshots | DSR: 0.95 -> 84.80 | Improved |
| apply LaSM scaling to layers 7-18 vs naive scaling of layers 21-26 (ablation) | Qwen2-VL-7B · Pop-up screenshots | DSR: 8.05 -> 84.80 | Improved |
| add LaSM to no-defense model | Qwen2-VL-7B · Overlay injection (avg) | DSR: 18.9 -> 66.4 | Improved |
| add LaSM to no-defense model | Qwen2-VL-7B · Inductive injection (avg) | DSR: 14.8 -> 68.3 | Improved |
| add LaSM to DPO-finetuned model | Qwen2-VL-7B · Inductive injection (avg) | DSR: 10.5 -> 65.2 | Improved |
| add LaSM to CoT alert | Qwen2-VL-7B · Inductive injection (avg) | DSR: 92.7 -> 99.8 | Improved |
| add LaSM to CoT alert | LLaVA-v1.6-Vicuna-13B · Overlay injection (avg) | DSR: 84.7 -> 86.9 | Mixed |
| tune scaling coefficient alpha from 1.00 to 1.10 (ablation) | LaSM on Qwen2-VL-7B · Pop-up screenshots | DSR: 1.42 -> 94.79 | Improved |
| tune scaling coefficient alpha from 1.00 to 1.20 (ablation) | LaSM on LLaVA-v1.6-Vicuna-13B · Pop-up screenshots | DSR: 50.24 -> 89.57 | Improved |
| add LaSM to model | Qwen2-VL-2B · Pop-up screenshots | DSR: 0.94 -> 23.20 | Improved |
| add LaSM to model | MAI-UI-8B · Pop-up screenshots | DSR: 15.17 -> 29.86 | Improved |
Closest papers by meaning
- Faithful Mobile GUI Agents with Guided Advantage Estimator (ICML 2026 · similarity 0.87)
- GhostEI-Bench: Do Mobile Agent Resilience to Environmental Injection in Dynamic On-Device Environments? (ICLR 2026 · similarity 0.86)
- AdapAction: Adaptive Target Action Backdoor Attack against GUI Agents (CVPR 2026 · similarity 0.86)
- GUI-G1: Understanding R1-Zero-Like Training for Visual Grounding in GUI Agents (NeurIPS 2025 · similarity 0.85)
- GTA1: GUI Test-time Scaling Agent (ICLR 2026 · similarity 0.85)
- SE-GUI: Enhancing Visual Grounding for GUI Agents via Self-Evolutionary Reinforcement Learning (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.
