Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

Retrieval metrics (recall@k, nDCG, MRR)

Retrieval metrics score how well a search system returns relevant documents. Recall@k checks how many relevant items appear in the top k, nDCG rewards ranking them high, and MRR measures how early the first correct one appears.

Also called: recall@k, nDCG, MRR, precision@k, mean reciprocal rank

By Divya Sharma · Checked Nov 15, 2026 · How we review

If your retrieval-augmented generation app gives bad answers, the first question is whether retrieval found the right passages. Retrieval metrics answer that separately from answer quality. You need a labelled set: queries with the documents that should come back, often part of a golden dataset.

Recall@k

Of all the relevant documents for a query, what share are in the top k results? If three passages answer a question and two are in the top 5, recall@5 is two thirds. For RAG this is often the most important number, since the model can only use what it is given.

Precision@k

Of the top k results, what share are relevant? Low precision means the prompt is full of noise.

nDCG

Normalized discounted cumulative gain uses graded relevance (very relevant, somewhat, not) and gives less credit the lower a good result appears. It is normalized to a 0 to 1 scale so queries can be compared. Many retrieval benchmarks, including the retrieval tasks in MTEB, report it.

MRR

Mean reciprocal rank looks only at the first correct result: 1 if it is first, 1/2 if second, 1/3 if third, averaged over queries. It suits tasks with one right answer.

Using them

Run the same query set whenever you change chunking, the embedding model, reranking or index settings. Pair retrieval metrics with LLM evaluation of the final answers.

Related terms

LLM evaluation
LLM evaluation, often called evals, is measuring how well a model or AI application performs a task by running it on a set of test cases and scoring the outputs against defined criteria, by code, by humans or by another model.
Golden dataset
A golden dataset is a fixed, carefully checked set of example inputs paired with the answers or behaviour you expect, used as the reference when you evaluate or regression-test an AI system.
MTEB
MTEB, the Massive Text Embedding Benchmark, is a public benchmark and leaderboard that scores text embedding models across many task types, such as retrieval, clustering, classification and reranking, over many datasets and languages.
Reranking
Reranking is a second search step that takes a short list of candidate documents and reorders them by relevance to the query, using a more accurate but slower model than the first search.
Retrieval-augmented generation
Retrieval-augmented generation (RAG) is a pattern where a system first searches a collection of documents for passages relevant to a question, then gives those passages to a language model so its answer draws on them.

Sources