RALM

A language model that incorporates a retrieval mechanism, fetching relevant documents from an external knowledge base as context before generating an answer. This approach reduces hallucinations and improves response accuracy by grounding generation in retrieved evidence.