HATS: Hardness-Aware Trajectory Synthesis for GUI Agents

TL;DR

HATS introduces a framework for generating more robust GUI agent trajectories by focusing on semantically ambiguous actions, improving generalization and alignment with instructions through iterative exploration and refinement.

Contribution

The paper presents a novel hardness-aware synthesis method that addresses semantic ambiguity in GUI actions, enhancing data quality for training more capable GUI agents.

Findings

Agents trained with HATS outperform baselines on benchmark environments.

HATS effectively identifies and mitigates semantic ambiguities in trajectory data.

The framework improves instruction-execution alignment in GUI tasks.

Abstract

Graphical user interface (GUI) agents powered by large vision-language models (VLMs) have shown remarkable potential in automating digital tasks, highlighting the need for high-quality trajectory data to support effective agent training. Yet existing trajectory synthesis pipelines often yield agents that fail to generalize beyond simple interactions. We identify this limitation as stemming from the neglect of semantically ambiguous actions, whose meanings are context-dependent, sequentially dependent, or visually ambiguous. Such actions are crucial for real-world robustness but are under-represented and poorly processed in current datasets, leading to semantic misalignment between task instructions and execution. To address these issues, we propose HATS, a Hardness-Aware Trajectory Synthesis framework designed to mitigate the impact of semantic ambiguity. We define hardness as the degree of semantic ambiguity associated with an action and develop two complementary modules: (1) hardness-driven exploration, which guides data collection toward ambiguous yet informative interactions, and (2) alignment-guided refinement, which iteratively validates and repairs instruction-execution alignment. The two modules operate in a closed loop: exploration supplies refinement with challenging trajectories, while refinement feedback updates the hardness signal to guide future exploration. Extensive experiments show that agents trained with HATS consistently outperform state-of-the-art baselines across benchmark GUI environments.