XSkill: Continual Learning from Experience and Skills in Multimodal Agents

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. Code is available at https://github.com/XSkill-Agent/XSkill.

AI assistants that analyze images and use tools like web search or small programs often fail to learn from past tasks: they repeat mistakes and use tools inefficiently. We built XSkill, which lets these systems improve without full retraining. It stores short tips from past attempts and broader step-by-step guides for common task types, using what it sees in images when learning and applying this knowledge. For each new task, XSkill finds relevant past advice, adapts it to the current question and images, and refines its knowledge over time. Across diverse image, search, and mixed tasks, XSkill outperformed tool-only systems and methods that learn from text alone. This could make such assistants more reliable, easier to audit, and better at applying lessons across tasks.