COVID-19 Scientific Article Collection for Ad-Hoc Search

TRECCOVID is a benchmark dataset containing scientific articles related to the COVID-19 pandemic, designed for evaluating ad-hoc search and text embedding models. It was created by the Massive Text Embedding Benchmark (MTEB) project and is hosted on Hugging Face, with a reference to the NIST COVID-19 information retrieval challenge. The dataset page was last updated in February 2026.

Use Cases

• Benchmarking embedding models on the TRECCOVID retrieval task using the MTEB framework.
• Evaluating search engine performance on COVID-19 scientific article queries.
• Training models to rank medical documents based on relevance to pandemic-related topics.

Strengths

• Dataset is part of the standardized Massive Text Embedding Benchmark (MTEB) for consistent model evaluation.
• Focuses on a specific, high-impact domain with articles from the COVID-19 pandemic.

Limitations

• Specific quantitative details like row count, column names, and dataset size are not provided in the input.
• The dataset's scope is limited to the COVID-19 pandemic, which may not generalize to other medical topics.

Provenance

Source

Based on the COVID-19 dataset from the NIST TREC COVID-19 information retrieval challenge (https://ir.nist.gov/covidSubmit/index.html).

Collection Method

Curated as part of the Massive Text Embedding Benchmark (MTEB) for evaluation tasks.

Time Range

Covers the period of the COVID-19 pandemic; specific years not stated.

Freshness

Dataset page was updated on 2026-02-24, indicating recent maintenance.

Geography

null