Tag
This paper introduces Agentic RAG-VLM, a unified framework that integrates retrieval-augmented generation with vision-language models and self-reflective planning for generalizable robotic grasping in cluttered environments, achieving 78.3% success rate.