DH{}DataHunter

Dataset from Hugging Face · anacodicai-labs

COVID-QA Retrieval Benchmark (AnacodicAI parse)

Open on Hugging Face
License
License not specified
The source does not state a license. Without one, the data may not be free to reuse: ask the author or pick another dataset.
Formats
JSON, MD, JSONL, CSV, ZST
Size
36 GB
Updated
Oct 9, 2026

Description

COVID-QA Retrieval Benchmark — parsed with the AnacodicAI guidelines pipeline A public retrieval / RAG benchmark over biomedical COVID-19 papers. Questions and gold answers come from deepset COVID-QA (Apache-2.0); the papers were fetched as PDFs and parsed with the same production pipeline as the AnacodicAI guidelines-generator and specialistRAG systems — so a number here reflects the real pipeline, not a toy one. 1,637 papers total: 137 with questions + 1,500 topic-matched…

Description from Hugging Face, shortened.

Files (8)

  • MDLICENSE.md771 byte
  • JSONLchunks.jsonl916 MB
  • JSONLgold/questions.jsonl539 kB
  • CSVmanifest.csv27 kB
  • CSVmanifest_distractors.csv253 kB
  • ZSTparsed_all_1637.tar.zst17 GB
  • ZSTparsed_decoys_1500.tar.zst16 GB
  • ZSTparsed_questions_137.tar.zst1.6 GB

DataHunter does not store or host these files: every link leads to the original source. Check the license there before using the data.

Search other datasets