CapeLLM: Support-Free Category-Agnostic Pose Estimation with Multimodal Large Language Models