Answer Accuracy
A metric in RAG system evaluation that measures whether an answer matches the annotated answer.
Timeline (last 90 days)
The experiments found that increasing the retrieval budget or switching to a larger retriever improves evidence recall, but this does not reliably translate into higher answer accuracy.
When evaluating the system, one should not look only at retrieval metrics such as recall@k, but should treat end-to-end answer accuracy, conflict-resolution correctness and refusal reasonableness as core monitoring metrics.
The three core evaluation metrics are answer accuracy, citation correctness and position sensitivity.
All Facts (3)
The experiments found that increasing the retrieval budget or switching to a larger retriever improves evidence recall, but this does not reliably translate into higher answer accuracy.
50%UnverifiedWhen evaluating the system, one should not look only at retrieval metrics such as recall@k, but should treat end-to-end answer accuracy, conflict-resolution correctness and refusal reasonableness as core monitoring metrics.
50%UnverifiedThe three core evaluation metrics are answer accuracy, citation correctness and position sensitivity.
50%