Neuro-Symbolic Skill Induction Boosts Agentic Task Performance

Researchers propose Neuro-Symbolic Skill Induction (NSI), a framework that transforms interaction traces into modular, logic-grounded programs. Unlike existing methods that produce state-blind scripts, NSI synthesizes explicit control flows and dynamic variable binding, enabling agents to learn when and why to act. The approach allows efficient generalization from few-shot examples and adaptation to unseen goals. Experiments on agentic tasks show NSI consistently outperforms state-of-the-art baselines, empowering agents to self-evolve.