Evaluating Fine-Tuned Embeddings for RAG: Methods & Code Snippets

Hello, Following the fine-tuning of the embedding model, are there any visual (like retrieved documents are closer to the question in 2d or something similar) comparison methods or code snippets available to assess how the fine-tuned model's embeddings better represent data, particularly in the context of RAG, in contrast to the embeddings generated by the base model?