Representation Learning With Large Language Models