Fundamentals

Retrieval Ranking

Retrieval ranking orders results by relevance. Rankers score document relevance to queries. Ranking can be lexical (keyword-based) or semantic (embedding-based). Learning-to-rank uses ML to optimize ranking. Good ranking puts most relevant documents first. Ranking quality affects downstream accuracy. Ranking efficiency matters for large document sets. Re-ranking (retrieving many, ranking top K) improves quality. Ranking is crucial for RAG system quality.