MS MARCO
A machine reading comprehension dataset released by Microsoft, built from Bing search queries and containing a large number of real questions with passage relevance annotations, commonly used for retrieval and question answering evaluation.
Timeline (last 90 days)
The unstructured retrieval field has widely recognized benchmarks such as MS MARCO and BEIR, whereas structured retrieval has long lacked a unified cross-domain evaluation set.
The industry commonly uses public benchmarks such as BEIR and MS MARCO to evaluate retrieval systems.
MS MARCO is a large-scale English question-answering dataset containing more than 500,000 real search queries and their relevant passage annotations, and has long served as a benchmark in the information retrieval field.
MTEB (Massive Text Embedding Benchmark) is currently the most widely cited text embedding evaluation leaderboard, covering tasks such as retrieval, classification, and clustering, and uses datasets including MS MARCO and BEIR.
All Facts (4)
The unstructured retrieval field has widely recognized benchmarks such as MS MARCO and BEIR, whereas structured retrieval has long lacked a unified cross-domain evaluation set.
50%UnverifiedThe industry commonly uses public benchmarks such as BEIR and MS MARCO to evaluate retrieval systems.
50%UnverifiedMS MARCO is a large-scale English question-answering dataset containing more than 500,000 real search queries and their relevant passage annotations, and has long served as a benchmark in the information retrieval field.
50%UnverifiedMTEB (Massive Text Embedding Benchmark) is currently the most widely cited text embedding evaluation leaderboard, covering tasks such as retrieval, classification, and clustering, and uses datasets including MS MARCO and BEIR.
50%