kandpal-2023-relevant-doc-both-entities-required
IN premise — summaries/2026/08/24/kandpal-2023-long-tail-knowledge-s3-lm-accuracy-depends-on-relevant.md
Created 2026-08-25T02:58:06+00:00
Counting a relevant document requires co-occurrence of both the salient question entity and the salient answer entity; counting either entity alone fails to predict accuracy when co-occurrence count is below 5.
Summary
A document only counts as relevant if it ties together both the key entity in the question and the key entity in the expected answer. Mentioning just one of those entities on its own is not a reliable signal of accuracy, especially when only a small number of documents meet the co-occurrence bar.