BoxLitE is a new knowledge base embedding model that utilizes convex optimization to enhance the representation of hierarchical concepts in knowledge graphs. This approach aims to improve the faithfulness of embeddings in knowledge representation tasks, potentially benefiting various applications in artificial intelligence and data management.