Definition
Learning-to-Rank (LTR) is a supervised machine learning approach used in the re-ranking stage of a RAG pipeline to optimize the ordering of retrieved documents, prioritizing precision in the LLM's context window at the cost of increased retrieval latency.
Focuses on the optimal sorting of retrieved results rather than the initial act of searching for them.
"A talent show judge scoring a pre-selected group of finalists to determine the gold, silver, and bronze winners."
Conceptual Overview
Learning-to-Rank (LTR) is a supervised machine learning approach used in the re-ranking stage of a RAG pipeline to optimize the ordering of retrieved documents, prioritizing precision in the LLM's context window at the cost of increased retrieval latency.
Disambiguation
Focuses on the optimal sorting of retrieved results rather than the initial act of searching for them.
Visual Analog
A talent show judge scoring a pre-selected group of finalists to determine the gold, silver, and bronze winners.