VT-FSL: Bridging Vision and Text with LLMs for Few-Shot Learning