← All questions
Retrieval & EmbeddingsMedium· Ranking· Aggregation· Bias

Why retrieve top-k per source instead of pooling every...

A weak answer treats the two aggregation schemes as equivalent. A strong answer identifies the specific failure of document-pooled voting: if a low-reliability source contributes many near-duplicate documents to the index, it can disproportionately influence the result. Per-source voting mitigates this by capping the influence of any single source to exactly one ballot, ensuring a more balanced representation.

Code playground (Python)