MS MARCO
MS MARCO pairs real, anonymised Bing search queries with passages that people have judged for relevance, giving a large and realistic benchmark for passage ranking and retrieval. It has become one of the standard reference points for measuring how well a system finds the right passage for a query.
The catch is the licence. MS MARCO is released under a non-commercial licence, and some model developers deliberately leave it out of their training data to avoid the restriction. If you are building anything commercial, read the terms before you rely on it.
Related sources
BEIR
A collection of 18 information retrieval datasets spanning ad-hoc web search, question answering, fact verification, and duplicate question retrieval. Built by aggregating existing datasets, some originally created for other tasks and converted to retrieval format. Tests how well retrieval models generalise to unseen domains without fine-tuning, which matters for real-world RAG.
BRIGHT
A reasoning-intensive retrieval benchmark. It requires genuine reasoning to connect queries with relevant documents rather than keyword or semantic overlap, revealing model weaknesses that BEIR misses.
CoIR (Code Retrieval Benchmark)
A code retrieval benchmark spanning diverse programming tasks and languages. Built for evaluating how well models retrieve relevant source code, which matters if you are building RAG over codebases.
LongEmbed
A retrieval benchmark focused on very long documents, with an average length above 5,500 words. It addresses a gap in standard benchmarks, which mostly use short documents of at most 512 tokens (the chunks of text a model reads at once).