Tag
This paper benchmarks Knowledge Graph-based augmentation against Retrieval Augmented Generation for culturally specific question answering, demonstrating significant error reduction on Latin American culture data using the LatamQA dataset.
The essay argues that literary disciplines provide essential tools for building culturally literate AI, proposing a layered framework to address structural monolingualism and promote pluralistic interpretations of AI textuality.