Learning-to-Rank

What is Learning-to-Rank?

Learning-to-rank (LTR) is supervised machine learning that optimizes the order of search results or recommendations from click logs and relevance labels—rather than hand-tuned heuristics alone.

In RAG and search assistants, ranking errors surface as wrong citations. Evaluate ranking metrics together with answer groundedness.

Related Giskard articles

• LLM hallucinations guide
• Giskard OSS v3

Evaluate ranked RAG answers

Combine LTR metrics with Giskard groundedness checks on top-k retrieved passages. Learn more.

Further reading

Authoritative reference: Wikipedia: Learning to rank.