Retrieval and QA (MS MARCO, Natural Questions, HotpotQA)
← Most-used data · All Text and NLP benchmarks datasets
Corpora and benchmarks for language understanding, question answering and text classification.
Tasks, access and usage conditions#
Mention counts below measure name matches in tracked papers, not benchmark quality or downloads. A family count can cover multiple datasets; it is not a separate count for each comparison row. Choose a task and version first, preserve evaluation splits, and follow the linked release terms. The site's content license does not replace the datasets' own terms.
| Dataset / version | Task and evaluation | Access and preparation | Usage conditions |
|---|---|---|---|
| MS MARCO | Passage/document retrieval, ranking and question answering using search questions. Evaluation: Use task-specific ranking or QA metrics and judgments; record corpus version and candidate-generation setup. | Official source / documentation Choose the specific task/release on the official page and accept its terms; ranking and QA collections differ. | Official terms restrict the datasets to non-commercial research; acceptance is required to access downloads. Source documents can carry separate rights. |
| Natural Questions | Questions from search with long/short answers annotated on Wikipedia pages. Evaluation: Official long-/short-answer metrics for full NQ; exact match for an open-domain derivative is a different task. | Official source / documentation Use the official release links and distinguish full NQ from simplified/open-domain derivatives. | The official repository is Apache-2.0 licensed; preserve the underlying Wikipedia content attribution and applicable terms as well. |
| HotpotQA | Multi-hop question answering with supporting-fact annotations. Evaluation: Answer EM/F1, supporting-fact scores and joint scores; retrieval difficulty differs sharply between the two settings. | Official source / documentation Download the official JSON release; choose distractor or fullwiki settings. | CC BY-SA 4.0 as stated by the creators; retain attribution and share-alike notices. |
Key findings#
- From Nov 2023 to Oct 2026, 68 archived papers mentioned Retrieval and QA (MS MARCO, Natural Questions, HotpotQA) in their title, abstract or keywords (the search of earlier months' full texts is still under way), most of them in Feb 2025 (5).
- No paper of the tracked journals mentioned Retrieval and QA (MS MARCO, Natural Questions, HotpotQA) in the last 30 days (2026-10-06 to 2026-11-04).
- From 2017 to 2026 (so far), Retrieval and QA (MS MARCO, Natural Questions, HotpotQA) was mentioned in 80 papers in the tracked journals, most of them in AI (75), information systems (3) and finance (1).
- 31 of them were published in 2025, against 0 in 2017.
- Since 2023, the institutions with the most of these papers were Tsinghua University (3), University of Isfahan (3) and Wuhan University (3).
- The journals publishing most of them were Knowledge-Based Systems (10), Artificial Intelligence Review (9) and Information Processing & Management (8).
- The most-cited of them is “Large language models in medical and healthcare fields: applications, advances, and challenges” (Artificial Intelligence Review, 2024), with 247 citations.
- Papers use Retrieval and QA (MS MARCO, Natural Questions, HotpotQA) most often with Wikipedia (30), Hugging Face Hub (17) and Twitter / X (9); it has 20 partners in the code and data network, in the group around Twitter / X, Facebook, YouTube / TikTok.
Mentions over 36 months#
| Trend | Nov 23–Oct 24 | Nov 24–Oct 25 | Nov 25–Oct 26 | 36 months |
|---|---|---|---|---|
| 11 | 27 | 30 | 68 |
Papers per year#
| Category | 2017 | 2018 | 2019 | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | 2026 (so far) |
|---|---|---|---|---|---|---|---|---|---|---|
| AI | 0 | 0 | 0 | 0 | 0 | 6 | 5 | 12 | 31 | 21 |
| Information Systems | 0 | 0 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 1 |
| Finance | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 1 |
| Economics | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 1 | 0 | 0 |
| All categories | 0 | 0 | 0 | 0 | 1 | 6 | 6 | 13 | 31 | 23 |
Where the papers come from#
| # | Institution | Papers |
|---|---|---|
| 1 | Tsinghua University | 3 |
| 2 | University of Isfahan | 3 |
| 3 | Wuhan University | 3 |
| 4 | Beijing Institute of Technology | 2 |
| 5 | Chinese Academy of Sciences | 2 |
| 6 | Eindhoven University of Technology | 2 |
| 7 | Loughborough University | 2 |
| 8 | Southeast University | 2 |
| 9 | Southern University of Science and Technology | 2 |
| 10 | Toronto Metropolitan University | 2 |
| # | Country | Papers |
|---|---|---|
| 1 | China | 28 |
| 2 | United Kingdom | 11 |
| 3 | United States | 7 |
| 4 | Australia | 6 |
| 5 | Italy | 6 |
| 6 | Canada | 5 |
| 7 | Hong Kong | 5 |
| 8 | Iran | 5 |
| 9 | Germany | 4 |
| 10 | Spain | 3 |
| # | Keyword | Papers |
|---|---|---|
| 1 | large language models | 32 |
| 2 | HotpotQA | 12 |
| 3 | retrieval-augmented generation | 11 |
| 4 | MS MARCO | 6 |
| 5 | dense retrieval | 6 |
| 6 | hallucination mitigation | 6 |
| 7 | information retrieval | 5 |
| 8 | knowledge graphs | 5 |
| 9 | multi-hop question answering | 4 |
| 10 | semantic search | 4 |
Most-cited papers#
Used together with#
| Used with | Type | Papers using both | Lift |
|---|---|---|---|
| Wikipedia | Dataset | 30 | 17.7 |
| Hugging Face Hub | Dataset | 17 | 21.8 |
| Twitter / X | Dataset | 9 | 1.2 |
| CoNLL | Dataset | 8 | 37.7 |
| GSM8K / MATH | Dataset | 8 | 90.3 |
| MMLU | Dataset | 8 | 60.6 |
| Dataset | 8 | 7.9 | |
| Dataset | 6 | 1.2 | |
| Natural language inference (SNLI, MultiNLI) | Dataset | 6 | 38.2 |
| BIG-bench | Dataset | 5 | 96.5 |
| SQuAD | Dataset | 5 | 115.0 |
| HumanEval / MBPP | Dataset | 4 | 70.3 |
Cite this page
Scholar4AI. (2026). Retrieval and QA (MS MARCO, Natural Questions, HotpotQA). Journal and Conference Research Rankings. Retrieved November 4, 2026, from https://scholar4ai.com/datasets/retrieval-and-qa-ms-marco-natural-questions-hotpotqa
@misc{scholar4ai-datasets-retrieval-and-qa-ms-marco-natural-questions-hotpotqa,
author = {{Scholar4AI}},
title = {Retrieval and QA (MS MARCO, Natural Questions, HotpotQA)},
year = {2026},
howpublished = {\url{https://scholar4ai.com/datasets/retrieval-and-qa-ms-marco-natural-questions-hotpotqa}},
urldate = {2026-11-04},
note = {Journal and Conference Research Rankings; data from OpenAlex (CC0); licensed CC BY-NC 4.0}
}