GUI-Perturbed: Domain Randomization Reveals Systematic Brittleness in GUI Grounding Models

TL;DR

GUI-Perturbed introduces a framework to evaluate GUI grounding models' robustness by systematically perturbing visual scenes and instructions, revealing vulnerabilities not captured by standard benchmarks.

Contribution

It provides a controlled perturbation framework, GUI-Perturbed, to diagnose specific weaknesses in GUI grounding models, including spatial reasoning and visual robustness.

Findings

Relational instructions cause systematic accuracy drops across models.

A 70% browser zoom significantly degrades model performance.

Fine-tuning with augmented data can worsen model accuracy.

Abstract

GUI grounding models report over 85% accuracy on standard benchmarks, yet drop 27-56 percentage points when instructions require spatial reasoning rather than direct element naming. Current benchmarks miss this because they evaluate each screenshot once with a single fixed instruction. We introduce GUI-Perturbed, a controlled perturbation framework that independently varies visual scenes and instructions to measure grounding robustness. Evaluating three 7B models from the same architecture lineage, we find that relational instructions cause systematic accuracy collapse across all models, a 70% browser zoom produces statistically significant degradation, and rank-8 LoRA fine-tuning with augmented data degrades performance rather than improving it. By perturbing along independent axes, GUI-Perturbed isolates which specific capability axes are affected-spatial reasoning, visual robustness, reasoning calibration-providing diagnostic signal that aggregate benchmarks cannot. We release the dataset, augmentation pipeline, and a fine-tuned model.