Search papers, labs, and topics across Lattice.
This paper critiques the limitations of current skill-learning methods that rely solely on text-based assets for agent capabilities, particularly in visual-centric tasks. It introduces a multimodal skill paradigm, \textbf{\NAME}, which integrates textual logic with visual support, allowing agents to leverage spatial layouts and visual grounding. Experimental results demonstrate that this approach significantly enhances performance in tasks requiring visual evidence and state-aware interactions, highlighting the necessity for agents to adopt multimodal skills.
Visual skills outperform text-only methods, revealing a critical gap in how agents learn and apply knowledge in complex, visual environments.
Reusable skills are a key mechanism for extending agent capabilities, allowing agents to accumulate experience and solve increasingly complex tasks. Yet most existing skill-learning methods store reusable experience as text-only assets, such as instructions, reasoning traces, or summarized trajectories. We argue that this text-only paradigm creates a fundamental bottleneck for visual-centric tasks, where reusable knowledge often depends on spatial layout, visual grounding, fine-grained appearance, and localized state changes. To address this limitation, we propose \textbf{\NAME}, a multimodal skill paradigm that combines declarative textual logic with explicit visual support. We distinguish three reusable forms: static priors for stable spatial conventions, dynamic priors for in-situ visual working memory, and interleaved visual skills that bind ordered text steps to the source frames, screenshots, or page regions that justify them. Rather than only describing what to do, visual skills also encode where to look, how to inspect, and how to verify visual outcomes. To scale visual-skill construction, we introduce \textbf{\SYSTEM}, an automatic system that converts agent experience into reusable multimodal skills by preserving textual reasoning, spatial references, visual boundaries, and interaction patterns from task trajectories. Experiments on GUI and other visual-centric tasks show that visual skills consistently outperform text-only skills, particularly when success requires spatial correspondence, visual evidence, and state-aware interaction. These results support our central position: reusable agent skills should go beyond text and become multimodal assets for future multimodal agents.
