XSkill: Continual Learning from Experience and Skills in Multimodal Agents

TL;DR

XSkill introduces a dual-stream continual learning framework for multimodal agents that leverages experiences and skills grounded in visual observations to improve tool use and reasoning without parameter updates.

Contribution

It proposes a novel dual-stream framework for continual learning from experience and skills, grounded in visual observations, enhancing multimodal agent reasoning and generalization.

Findings

XSkill outperforms baselines across five benchmarks.

The two knowledge streams complement each other in reasoning.

XSkill shows superior zero-shot generalization.

Abstract

Multimodal agents can now tackle complex reasoning tasks with diverse tools, yet they still suffer from inefficient tool use and inflexible orchestration in open-ended settings. A central challenge is enabling such agents to continually improve without parameter updates by learning from past trajectories. We identify two complementary forms of reusable knowledge essential for this goal: experiences, providing concise action-level guidance for tool selection and decision making, and skills, providing structured task-level guidance for planning and tool use. To this end, we propose XSkill, a dual-stream framework for continual learning from experience and skills in multimodal agents. XSkill grounds both knowledge extraction and retrieval in visual observations. During accumulation, XSkill distills and consolidates experiences and skills from multi-path rollouts via visually grounded summarization and cross-rollout critique. During inference, it retrieves and adapts this knowledge to the current visual context and feeds usage history back into accumulation to form a continual learning loop. Evaluated on five benchmarks across diverse domains with four backbone models, XSkill consistently and substantially outperforms both tool-only and learning-based baselines. Further analysis reveals that the two knowledge streams play complementary roles in influencing the reasoning behaviors of agents and show superior zero-shot generalization.