EvolveNav: Proactive Preflection and Self-Evolving Memory for Zero-Shot Object Goal Navigation

The paper introduces EvolveNav, a self-evolving framework for Zero-Shot Object-Goal Navigation (ZS-OGN) that enhances embodied agents' ability to locate target objects without prior training. Key innovations include an agentic rule memory that captures actionable knowledge from past experiences and a memory-guided preflection module that predicts outcomes to minimize inefficient exploration. The proposed method demonstrates a 10.1% improvement in success rate over existing zero-shot baselines, making it significant for practitioners seeking to improve navigation efficiency in AI systems.