MS MARCO is a collection of datasets focused on deep learning in search, part of the Massive Text Embedding Benchmark (MTEB). The dataset was last updated on 2025-05-04 and is hosted by the mteb organization on Hugging Face. It covers domains including Encyclopaedic, Academic, Blog, News, Medical, Government, Reviews, Non-fiction, Social, and Web content.
Use Cases
- Benchmarking text embedding models based on the dataset's role in the Massive Text Embedding Benchmark.
- Training search relevance models based on the dataset's focus on deep learning in search.
- Evaluating model performance across multiple domains based on the listed domains such as Academic, Medical, and News.
Strengths
- Part of the Massive Text Embedding Benchmark (MTEB), a standard for evaluating embeddings.
- Covers a wide range of domains including Academic, Medical, and Government content as listed in the description.
- Last updated on 2025-05-04, indicating recent maintenance.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- mteb (via Hugging Face), referencing https://microsoft.github.io/msmarco/.
- Collection Method
- Likely aggregated from multiple web and document sources for search tasks.
- Time Range
- null
- Freshness
- Last updated 2025-05-04 16:10:23.
- Geography
- null