Explainability is critical when we are modeling data. This is no less true for GenAI. This chart shows the group of embeddings that OpenAI used to answer the question: "Who attacked Israel in 2023?" This is the support for the answer based on additional information given to the model.
