datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
lean-rag-indexes
Lean RAG Indexes for BioASQ
Paper: Retrieval-Bound Generation: Lean RAG Pipelines for Biomedical QA — CLEF 2026 Working Notes, BioASQ Task 14b
Code: github.com/lasigeBioTM/BioASQ14Taskb_2026
Prebuilt retrieval indexes for the Lean RAG Pipelines for Biomedical Question Answering project, developed as part of an MSc dissertation at LASIGE, University of Lisbon (in preparation).
These indexes support a hybrid (BM25 + dense retrieval) pipeline evaluated on BioASQ Task 14b.… See the full description on the dataset page: https://huggingface.co/datasets/dantunes6/lean-rag-indexes.GeneratingQuestions
HVU_QA
HVU_QA is an open-source Vietnamese Question-Context-Answer (QCA) corpus, accompanied by supporting tools, created to facilitate the development of FAQ-style question generation and question answering systems, particularly for low-resource language settings. The dataset was developed by a research team at Hung Vuong University, Phu Tho, Vietnam, led by Dr. Ha Nguyen, Deputy Head of the Department of Engineering Technology. HVU_QA was constructed using a fully automated… See the full description on the dataset page: https://huggingface.co/datasets/DANGDOCAO/GeneratingQuestions.ckg-benchmark
CKG Benchmark
Pre-structured knowledge graphs outperform RAG by 4× F1 at 11× lower token cost — across 65 benchmarked domains.
System
Macro F1
Tokens/query
RDS
Run Cost
CKG
0.4709
269
0.00175
$7.81
RAG
0.1231
2,982
0.0000413
$76.23
GraphRAG
0.1200
3,450
0.0000452
$44.43
42× more intelligence per token than RAG. Auditable by design — every answer traces to a declared edge.
Dataset Contents
domains/{domain}/learning-graph.csv — structured DAG… See the full description on the dataset page: https://huggingface.co/datasets/danyarm/ckg-benchmark.norwegian-dyna-instruct
🧨 Norwegian dyna-instruct
Version
0.1.0 (changelog)
Languages
Norwegian Bokmål (nob), Norwegian Nynorsk (nno), and English (eng) translation input
License
Mixed open licenses; see the table below
Sources
Five datasets (source cards)
Dataset Description
Number of samples: 14.40K
Number of tokens (Llama 3): 6.27M
Average conversation length in tokens (min, max): 435.63 (4, 8.92K)
Average number of turns (min, max): 2.13 (2, 3)… See the full description on the dataset page: https://huggingface.co/datasets/danish-foundation-models/norwegian-dyna-instruct.MMFineReason-1.8M-Qwen3-VL-235B-Thinking
MMFineReason
Closing the Multimodal Reasoning Gap via Open Data-Centric Methods
Average score across mathematical reasoning and multimodal understanding benchmarks.
📖 Overview
MMFineReason is a large-scale, high-quality multimodal reasoning dataset comprising 1.8M samples and 5.1B solution tokens, featuring detailed reasoning annotations distilled from Qwen3-VL-235B-A22B-Thinking.
🎯 Key Highlights
1.8M High-Quality Samples with 5.1B Solution Tokens… See the full description on the dataset page: https://huggingface.co/datasets/dans25275/MMFineReason-1.8M-Qwen3-VL-235B-Thinking.wikitablequestions-wtq
Dataset Card for WikiTableQuestions-wtq
Dataset Summary
The WikiTableQuestions-wtq dataset is a small-scale dataset for the task of question answering on semi-structured tables.
This data includes the aggregation_label and answer_coordinates to make it easy to train this model on any TAPAS based modles.
Supported Tasks and Leaderboards
question-answering, table-question-answering
Languages
en
Dataset Structure
Data… See the full description on the dataset page: https://huggingface.co/datasets/danwakeem/wikitablequestions-wtq.alpaca-cleaned-italian
Dataset Card for Alpaca-Cleaned-Italian
About the translation and the original data
The translation was done with X-ALMA, a 13-billion-parameter model that surpasses state-of-the-art open-source multilingual LLMs (as of Q1 2025, paper here).
The original alpaca-cleaned dataset is also kept here so that there is parallel data for Italian and English.
Additional notes on the translation
Despite the good quality of the translation, errors, though rare, are… See the full description on the dataset page: https://huggingface.co/datasets/DanielSc4/alpaca-cleaned-italian.attention-uq-800q-colab
Attention/UQ 800-question Colab bundle
A deterministic 200-question subset for each of MultiModalQA, WebQA, HotpotQA, and TAT-QA. See manifest.json for exact upstream sources, hashes, counts, and the explicitly constructed WebQA distractor setting.
OpenCodeReasoningRubrics
OpenCodeReasoningRubrics
This dataset contains questions and rubric annotations intended for evaluating reasoning quality in open-ended coding or logical tasks.
Dataset Structure
Each example in the dataset has the following fields:
index (int): A unique identifier.
question (str): A natural language question or prompt.
rubric (str): An explanation or rubric detailing expectations or evaluation criteria.
The data is stored in split .parquet files for efficient loading… See the full description on the dataset page: https://huggingface.co/datasets/danikhan632/OpenCodeReasoningRubrics.RubricHub_v1
RubricHub
RubricHub is a large-scale (approximately 110K), multi-domain dataset that provides high-quality rubric-based supervision for open-ended generation tasks. It is constructed via an automated coarse-to-fine rubric generation framework, which integrates principle-guided synthesis, multi-model aggregation, and difficulty evolution to produce comprehensive and highly discriminative evaluation criteria, overcoming the supervision ceiling of… See the full description on the dataset page: https://huggingface.co/datasets/dans25275/RubricHub_v1.Dans-MemoryCore-CoreCurriculum-Small
Dan's Memory Core: Core Curriculum Small
Broad strokes
This dataset aims to provide a foundation of knowledge common to a number of fields and areas of study. The question answer pairs were generated using a RAG implementation and a curated selection of source material. Ideally this will be the first in a series of datasets that will cover a wide range of topics.
Nomic Atlas Visualiztion
Cluster visualization for the dataset available here.
Topics… See the full description on the dataset page: https://huggingface.co/datasets/PocketDoc/Dans-MemoryCore-CoreCurriculum-Small.mmqa-first100-colab
MultiModalQA first-100 Colab subset
This repository contains the first 100 examples of the official MultiModalQA dev split and only their referenced text, table, and image assets. It is a reproducibility artifact for Untitled34_attention_uq_100q_benchmark.ipynb.
The original dataset is from allenai/multimodalqa. See manifest.json for counts and source hashes.
Med-Evidence-2.6k
Med-Evidence-2.6k
Med-Evidence-2.6k is a benchmark for evaluating evidence-grounded medical diagnostic reasoning. It was developed as part of the work EviDx: Evidence-Aware Active Diagnosis with Scaffolded LLM Agents, accepted to Findings of EMNLP 2026.
The benchmark provides clinical diagnostic questions paired with ground-truth answers and annotated evidence spans, enabling evaluation of both diagnostic accuracy and evidence-grounded reasoning.
Dataset Structure… See the full description on the dataset page: https://huggingface.co/datasets/danieez/Med-Evidence-2.6k.EgoDynamic4D
EgoDynamic4D Dataset from AAAI 2026 paper: Understanding Dynamic Scenes in Egocentric 4D Point Clouds
🚀 EgoDynamic4D QA dataset has been released.
This repository hosts the official implementation of EgoDynamic4D, a large-scale egocentric 4D dynamic scene understanding benchmark introduced in our AAAI 2026 paper:
Understanding Dynamic Scenes in Egocentric 4D Point Clouds
About the Dataset
EgoDynamic4D is a question answering (QA) benchmark designed for… See the full description on the dataset page: https://huggingface.co/datasets/Dancing-Huggingface/EgoDynamic4D.eli5_question_answer_danish
ELI5 question-answer pairs in Danish
About
This dataset is a version of the ELI5 question-answer pairs dataset machine-translated from English to Danish (link to original dataset).
Machine translation is performed using the Helsinki NLP English-to-Danish OPUS-MT model.
The dataset contains ~209k question-answer pairs and can be used to train embedding and question-answer models. Each pair consists of one question ('query') and one passage containing the answer… See the full description on the dataset page: https://huggingface.co/datasets/KennethTM/eli5_question_answer_danish.Danish-llama-gen
Dataset
Experimental dataset generation using LLaMa-3-70B.
The dataset was created using same techniques as Stanford-Alpaca, but using Open-Source LLM's.
This was generated using only 170 seed tasks and shows a big potential in domain-specific data generation, even for Danish.
Limitations
The generations are very basic and short form, mainly due to the prompt-template. English text has been sorted out but sometimes confuses Danish/Swedish/Norwegian words.
This should… See the full description on the dataset page: https://huggingface.co/datasets/Mabeck/Danish-llama-gen.mot-dang-chiang-mai-chiang-rai
มดแดง Mot Dang — Chiang Mai & Chiang Rai city directory
88,161 places in and around Chiang Mai (62,772) and Chiang Rai (25,389),
in Thai and English, with coordinates, categories, opening hours, and the
channels a place actually answers on — phone, LINE, Facebook, a website that
still resolves.
The name is มดแดง, mot daeng, the red ant: the thing that knows every soi
because it has walked all of them. That is the ambition. The directory exists
because mainstream mapping is thin… See the full description on the dataset page: https://huggingface.co/datasets/NaNoBotCo/mot-dang-chiang-mai-chiang-rai.mquad-v1
MQuAD
The Medical Question and Answering dataset(MQuAD) has been refined, including the following datasets. You can download it through the Hugging Face dataset. Use the DATASETS method as follows.
Quick Guide
from datasets import load_dataset
dataset = load_dataset("danielpark/MQuAD-v1")
Medical Q/A datasets gathered from the following websites.
eHealth Forum
iCliniq
Question Doctors
WebMD
Data was gathered at the 5th of May 2017.
The MQuAD provides embedded question… See the full description on the dataset page: https://huggingface.co/datasets/danielpark/mquad-v1.danish-OpenHermes
Dataset
This is a translated version of a subset from OpenHermes. Coding tasks and word-play such as anagrams have been removed.
It has been translated using SeamlessM4T v2 T2T.
SUPERChem
SUPERChem: A Multimodal Reasoning Benchmark in Chemistry
🌐 Website | 📄 Paper | 💻 Code
📢 Updates
[2025-12-06] PDF Preview Released: We have released the PDF version of SUPERChem in both English and Chinese to facilitate easier previewing and manual inspection, especially for non-technical users. You can download SUPERChem-500.zip to access the dataset in PDF format. The password to unzip the file is SUPERChem2025.
🧪 What is SUPERChem?
SUPERChem is… See the full description on the dataset page: https://huggingface.co/datasets/DanielW11/SUPERChem.tt
Ultimate Red Team AI Training Dataset 💀
Dataset Description
A comprehensive dataset for training AI models in offensive security, red team operations, and penetration testing. This dataset combines real-world vulnerability data, exploitation techniques, and operational frameworks to create an AI capable of autonomous red team operations.
Dataset Summary
Total Data Points: 550,000+ unique security-related entries
Categories: 15+ major security domains… See the full description on the dataset page: https://huggingface.co/datasets/DannySc/tt.sparql-dataset-era-64k
Contenido
Este dataset contiene parejas de consultas en lenguaje natural y sus equivalentes aproximados en SPARQL, generados automáticamente usando Qwen3-30B alimentado con fragmentos de la ontología de ERA (v3.0) así como valores reales contenidos en los grafos de RINF y ERATV.
Se ha alimentado la generación con instancias concretas, conjuntos de clases y propiedades, así como valores de SKOS.
Proceso
Se generaron 64K muestras de prompts para emplear por un LLM.
Se… See the full description on the dataset page: https://huggingface.co/datasets/daniel-dona/sparql-dataset-era-64k.African-History-QA-Dataset
Dataset Name
African History Dataset
Dataset Structure
Data Fields
question: Questions about African History
answer: Answers to Questions.
Data Splits
Training: 2114 examples
Validation: 200 examples
Testing: 100 examples
Usage
from datasets import load_dataset
dataset = load_dataset("DannyAI/African-History-QA-Dataset")
Citation Information
If you use this dataset, please cite:
@dataset{
Ihenacho2026African_History_Dataset… See the full description on the dataset page: https://huggingface.co/datasets/DannyAI/African-History-QA-Dataset.Danish-SlimOrca
Dataset
A subset of the SlimOrca dataset translated using Meta-Llama-3-70B-Instruct.
Retro-YahooAnswers
Description
This dataset is an instruct style dataset comprised of a scrape of the Yahoo! Answers website that was done in 2007. The dataset is comprised of 10 categories labeled 1-10. The categories are as follows:
Society & Culture
Science & Mathematics
Health
Education & Reference
Computers & Internet
Sports
Business & Finance
Entertainment & Music
Family & Relationships
Politics & Government
The subject line and body of the question have been combined into a single field and… See the full description on the dataset page: https://huggingface.co/datasets/Dans-DiscountModels/Retro-YahooAnswers.gooaq_pairs_danish
GooAQ (Google Answers to Google Questions) question-answer pairs in Danish
About
This dataset is a version of the GooAQ question-answer pairs dataset machine-translated from English to Danish (link to original dataset).
Machine translation is performed using the Helsinki NLP English-to-Danish OPUS-MT model.
The dataset contains ~3M question-answer pairs and can be used to train embedding and question-answer models. Each pair consists of one question ('query') and one… See the full description on the dataset page: https://huggingface.co/datasets/KennethTM/gooaq_pairs_danish.Dans-Assistantmaxx-Synthia
Source:
@misc {agicommies_2024,
author = { {agicommies} },
title = { synthia (Revision 914b306) },
year = 2024,
url = { https://huggingface.co/datasets/agicommies/synthia },
doi = { 10.57967/hf/2125 },
publisher = { Hugging Face }
}
ragu-webq-contriever-msmarco
RAGU WebQuestions retrieval data
This repository contains WebQuestions train/dev/test splits prepared for the
RAGU codebase. Each record contains question, answers, q_id, and ctxs.
Retrieval provenance
Retriever: facebook/contriever-msmarco
Similarity: exact flat inner product
Knowledge base: DPR 2018 English Wikipedia, 100-word passages
Retrieved contexts retained per question: 20
WebQ source: official DPR resources
Retrieval was performed shard by shard to… See the full description on the dataset page: https://huggingface.co/datasets/danny2507/ragu-webq-contriever-msmarco.danish-citizenzhip-test-mcq
Dataset Card for "danish-citizen-test-mcq"
Dataset Description
This dataset covers Danish tests for both citizenship ("indfødsretsprøven") and permanent residence ("medborgerskabsprøven"), from 2016-2024.
Data follows the Aya Expedition format for global exams. Only unique questions between exams are kept.
Languages
The dataset is available in Danish (da).
Dataset Structure
An example from the dataset looks as follows.
{
"language": "da",
"country":… See the full description on the dataset page: https://huggingface.co/datasets/tellarin-ai/danish-citizenzhip-test-mcq.danoliterate-survey-prompts
Danoliterate Survey Prompts
Introduction
Outputs from 18 different LLM's on 100 prompts each based on of 100 popular use-cases of Generative AI
Data used for the Danoliterate Survey with Human Feedback on Generative, Large Language Models in Danish.
If you speak Danish, add your own response: danoliterate.compute.dtu.dk/Spørgeskema
For more details, see danoliterate.compute.dtu.dk/Articles