The author proposes Domain-Agnostic Mutual Prompting (DAMP) to align visual and textual embeddings for domain-invariant representations in Unsupervised Domain Adaptation (UDA).