Instruction: Explain why cited evidence improves a retrieval-based assistant.
Context: Checks whether the candidate can explain the core concept clearly and connect it to real production decisions. Explain why cited evidence improves a retrieval-based assistant.
The way I'd think about it is this: Citations matter because they turn an answer from something that sounds right into something that can be checked quickly. In a RAG system, the user is not just asking for fluency. They are asking whether the claim came from a real source they can inspect.
They also matter operationally. When an answer is wrong, citations help separate retrieval failures from synthesis failures. If I cannot see what evidence the model used, debugging gets expensive fast because every problem looks like a vague model issue.
I also like citations because they create a healthier product contract. The assistant does not have to pretend omniscience. It can show the governing evidence, acknowledge when support is partial, and make abstention feel legitimate instead of awkward. The bar is not simply attaching a source name. Good citations should map to the actual claim and point to the right document version.
A weak answer is saying citations are mostly for compliance or to make the UI look trustworthy. That ignores their real value for debugging, claim verification, and teaching the system to fail honestly.
easy
easy
easy
easy
easy
easy