NEXUS: Continual Learning of Symbolic Constraints for Safe and Robust Embodied Planning
For embodied AI systems, this work addresses the critical gap between probabilistic LLM outputs and the deterministic safety requirements of physical world tasks.
NEXUS introduces a modular framework for continual learning in embodied agents that decouples physical feasibility from safety specifications, achieving superior task success rates and effectively refusing unsafe instructions on SafeAgentBench while progressively improving planning efficiency.
While Large Language Models (LLMs) have catalyzed progress in embodied intelligence, a fundamental gap between their inherent probabilistic uncertainty and the strict determinism and verifiable safety required in the physical world. To mitigate this gap, this paper introduces NEXUS, a modular framework designed for continual learning in embodied agents. Different from prior works that treat symbolic artifacts merely as static interfaces, NEXUS leverages them for symbolic grounding and knowledge evolution. The framework explicitly decouples physical feasibility from safety specifications: capability of agents is improved through closed-loop execution feedback, while probabilistic risk assessments are grounded into deterministic hard constraints to establish a rigorous pre-action defense. Experiments on SafeAgentBench demonstrate that NEXUS achieves superior task success rates while effectively refusing unsafe instructions, exhibiting robust defense against adversarial attacks, and progressively improving planning efficiency through knowledge accumulation.