datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
casimedicos-exp
Antidote CasiMedicos Dataset - Possible Answers Explanations in Resident Medical Exams
We present a new multilingual parallel medical dataset of commented medical exams which includes not only explanatory arguments
for the correct answer but also arguments to explain why the remaining possible answers are incorrect.
This dataset can be used for various NLP tasks including: Medical Question Answering, Explanatory Argument Extraction or Explanation Generation.
The… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/casimedicos-exp.MedExpQA
MexExpQA: Multilingual Benchmarking of Medical QA with reference gold explanations and Retrieval Augmented Generation (RAG)
We present a new multilingual parallel medical benchmark, MedExpQA, for the evaluation of LLMs on Medical Question Answering.
This benchmark can be used for various NLP tasks including: Medical Question Answering or Explanation Generation.
Although the design of MedExpQA is independent of any specific dataset, for the first version of the… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/MedExpQA.YiZhaoEusExams
Dataset Card for EusExams
[!WARNING]
A newer version of this dataset is available! Please use EusExams-v2 which features deduplication, data grouping, and new data.
EusExams is a collection of tests designed to prepare individuals for Public Service examinations conducted by several Basque institutions, including the public health system Osakidetza, the Basque Government, the City Councils of Bilbao and Gasteiz, and the University of the Basque Country (UPV/EHU). Within each… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/EusExams.BertaQA
Dataset Card for BertaQA
BertaQA is a trivia dataset comprising 4,756 multiple-choice trivia questions, with one single correct answer and 2 additional distractors. Crucially, questions are distributed between local and global topics. Whereas answering questions in the latter group requires general world knowledge, local questions require specific knowledge about the Basque Country and its culture. Additionally, questions are classified into eight categories, namely Basque and… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/BertaQA.latxa-corpus-v1.1
Latxa Corpus v1.1
This is the training corpus of Latxa v1.1, a family of large language models for Basque based on Llama 2.
💻 Repository: https://github.com/hitz-zentroa/latxa
📒 Blog Post: Latxa: An Open Language Model and Evaluation Suite for Basque
📖 Paper: Latxa: An Open Language Model and Evaluation Suite for Basque
📧 Point of Contact: hitz@ehu.eus
📌 Notice
As of February 13th 2026, this repository reflects a curated version of the original dataset.
Some data… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/latxa-corpus-v1.1.ars-magna-greatest-hits
Ars Magna Greatest Hits
The funniest and most apt anagrams of people, companies, products, titles, places and phrases, found by Ars Magna and kept by hand.
Every row is a real anagram: the words use exactly the input's letters, checked against a pinned revision of English OpenList (368bf0e4460461c985fca8bde49e4062d56c1516), and every word is in the tier the row names. Accented letters fold to their base letter, so Beyoncé has three e's. Nothing typed is ever replaced by… See the full description on the dataset page: https://huggingface.co/datasets/ryanjosephkamp/ars-magna-greatest-hits.latxa-corpus-v2
Latxa Corpus v2
📧 Point of Contact: hitz@ehu.eus
Dataset Summary
Curated by: HiTZ Research Center & IXA Research group (University of the Basque Country UPV/EHU)
Language(s): eu-ES
Latxa Corpus v2 is a large-scale monolingual Basque corpus, created by combining curated crawls, public datasets, institutional data, and newly collected resources.
Compared to v1.1, it substantially increases coverage, diversity, and volume.
The final corpus is deduplicated, filtered, and… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/latxa-corpus-v2.PRMBench_Preview🏠 PRM-Eval Homepage | 💻 Code | 📑 Paper | 📚 PRM Eval Documentation
Introduction
This is the official dataset for PRMBench. PRMBench is a benchmark dataset for evaluating process-level reward models (PRMs). It consists of 6,216 data instances, each containing a question, a solution process, and a modified process with errors. The dataset is designed to evaluate the ability of PRMs to identify fine-grained error types in the solution process. The dataset is annotated with error… See the full description on the dataset page: https://huggingface.co/datasets/hitsmy/PRMBench_Preview.EusProficiency
Dataset Card for EusProficiency
EusProficiency comprises 5,169 exercises on different topics from past EGA exams, the official C1-level certificate of proficiency in Basque.
We collected the atarikoa exercises from EGA exams through the years 1998 to 2008. Atarikoa is the first qualifying test of EGA, which measures different aspects of language competency, such as reading comprehension, grammar, vocabulary, spelling, and writing. Each test generally has 85 multiple-choice questions… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/EusProficiency.EusReading
Dataset Card for EusReading
EusReading consists of 352 reading comprehension exercises (irakurmena) sourced from the set of past EGA exams from 1998 to 2008. Each test generally has 10 multiple-choice questions, with 4 choices and a single correct answer. These exercises are more challenging than Belebele due to the complexity and length of the input texts. As a result, EusReading is useful to measure long context understanding of models.
Curated by: HiTZ Research Center & IXA… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/EusReading.truthfulqa-multi
Dataset Card for TruthfulQA-multi
TruthfulQA-multi is a professionally translated extension of the original TruthfulQA benchmark designed to evaluate truthfulness in Basque, Catalan, Galician, and Spanish. The dataset enables evaluating the ability of Large Language Models (LLMs) to maintain truthfulness across multiple languages.
Dataset Details
Dataset Description
TruthfulQA-multi extends the original English TruthfulQA dataset to four additional languages… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/truthfulqa-multi.ARC-eu
Dataset Card for ARC-eu
Point of Contact: hitz@ehu.eus
Dataset Description
Dataset Summary
ARC-eu is the professional translation to Basque of ARC's
(Clark et al., 2018) validation and test partitions.
ARC is a QA benchmark of grade-school level, multiple-choice science questions.
Languages
eu-ES
Dataset Structure
Data Instances
ARC-eu examples look like this:
{
"id": "MCAS_2000_4_6",
"question": "Zein teknologia… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/ARC-eu.BizFinBench
BizFinBench: A Business-Driven Real-World Financial Benchmark for Evaluating LLMs
📖Paper |🐙Github|🤗Huggingface
Large language models excel in general tasks, yet assessing their reliability in logic‑heavy, precision‑critical domains like finance, law, and healthcare remains challenging. To address this, we introduce BizFinBench, the first benchmark specifically designed to evaluate LLMs in real-world financial applications. BizFinBench comprises over 100,000+ bilingual (English &… See the full description on the dataset page: https://huggingface.co/datasets/HiThink-Research/BizFinBench.EusTrivia
Dataset Card for EusTrivia
EusTrivia consists of 1,715 trivia questions from multiple online sources. 56.3% of the questions are elementary level (grades 3-6), while the rest are considered challenging. A significant portion of the questions focus specifically on the Basque Country, its language and culture. Each multiple-choice question contains two, three or four choices (3.84 on average) and a single correct answer. Five areas of knowledge are covered:
Humanities and Natural… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/EusTrivia.MGSM-eu
Dataset Card for MGSM-eu
Point of Contact: hitz@ehu.eus
Dataset Description
Dataset Summary
MGSM (Shi et al., 2023) is a subset of 250 grade-school math problems from the GSM8K dataset (Cobbe et al., 2021) that has been manually translated into 10 typologically diverse languages.
Here, we provide professional translations to yet another language: Basque.
Languages
eu-ES
Dataset Structure
Data Instances
MGSM-eu train examples… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/MGSM-eu.hitchcock-psycho-1960-film-dataset-transformed
Psycho → AI-Model Dataset (Transformed)
A thematic re-skin of the Psycho (1960) Q&A dataset into an original AI-model setting where the world is transformed into an AI/data-center environment.
Character names, actor names, objects, locations, production references, dates, and thematic elements are remapped to AI/ML concepts and modern technology.
File: psycho_dataset_transformed.jsonl
Format: JSONL — one JSON object per line
Schema: each line has prompt and completion string… See the full description on the dataset page: https://huggingface.co/datasets/antfr99/hitchcock-psycho-1960-film-dataset-transformed.Multilingual-Opinion-Target-ExtractionThis repository contains the English 'SemEval-2014 Task 4: Aspect Based Sentiment Analysis'. translated with DeepL into Spanish, French, Russian, and Turkish. The labels have been manually projected. For more details, read this paper: Model and Data Transfer for Cross-Lingual Sequence Labelling in Zero-Resource Settings.
Intended Usage: Since the datasets are parallel across languages, they are ideal for evaluating annotation projection algorithms, such as T-Projection.
Label… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/Multilingual-Opinion-Target-Extraction.Hi-ToM_Dataset
Hi-ToM Dataset
This is the dataset for the paper "Hi-ToM: A Benchmark for Evaluating Higher-Order Theory of Mind Reasoning in Large Language Models".
The Hi-ToM_data folder
Contains ToMh data consisting of story-question pairs and the corresponding answers.
The names of subfolder branches have the following meanings:
Tell / No_Tell: whether or not the stories contain communications among agents.
MC / CoT: the prompting style. MC corresponds to Vanilla Prompting (VP) in… See the full description on the dataset page: https://huggingface.co/datasets/Hi-ToM/Hi-ToM_Dataset.hitabmeta4xnli
Dataset Card for Dataset Name
Meta4XNLI is a parallel dataset with annotations in English and Spanish for metaphor detection at token level (13320 sentences) and metaphor interpretation framed within NLI the task (9990 premise-hypothesis pairs).
It is a collection of existing NLI datasets manually labeled for both metaphor tasks.
Repository: data available also in .tsv format at https://github.com/elisanchez-beep/meta4xnli
Paper: Meta4XNLI: A Crosslingual Parallel Corpus for… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/meta4xnli.PIQA-eu
Dataset Card for PIQA-eu
Point of Contact: hitz@ehu.eus
Dataset Description
Dataset Summary
PIQA-eu is the professional translation to Basque of the PIQA's
(Bisk et al., 2020) validation partition.
PIQA is a commonsense QA benchmark for naive physics reasoning focusing on how we interact with everyday
objects in everyday situations.
Languages
eu-ES
Dataset Structure
Data Instances
PIQA-eu examples look like this:
{… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/PIQA-eu.XCOPA-eu
Dataset Card for XCOPA-eu
Point of Contact: hitz@ehu.eus
Dataset Description
Dataset Summary
XCOPA-eu is the professional translation to Basque of the English COPA dataset (Roemmmele et al., 2011),
in the spirit of the XCOPA effort (Ponti et al., 2020).
COPA is a dataset of causal commmonsense reasoning that focuses on cause-effect relationships between a
premise and two choices.
Languages
eu-ES
Dataset Structure… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/XCOPA-eu.Multilingual_MMEThis is the multilingual version of MME. If you find it useful, we hope you can cite our paper.
@article{ye2025claim,
title={CLAIM: Mitigating Multilingual Object Hallucination in Large Vision-Language Models with Cross-Lingual Attention Intervention},
author={Ye, Zekai and Li, Qiming and Feng, Xiaocheng and Qin, Libo and Huang, Yichong and Li, Baohang and Jiang, Kui and Xiang, Yang and Zhang, Zhirui and Lu, Yunfei and others},
journal={arXiv preprint arXiv:2506.11073},
year={2025}
}
CIGEval_sft_data
Dataset Card for CIGEval_sft_data
CIGEval_sft_data is the dataset used for fine-tuning LMMs in the paper CIGEval. It contains data on both tool selection and image evaluation, which can be combined into 2.3k complete evaluation trajectories. The dataset was constructed through the following steps:
Using GPT-4o + CIGEval to evaluate the full ImageHub dataset, generating 4,903 evaluation trajectories.
Randomly selecting 60% of these and filtering out the ones where the evaluation… See the full description on the dataset page: https://huggingface.co/datasets/HIT-TMG/CIGEval_sft_data.elkarhizketak-RAG
Dataset Card for ElkarHizketak RAG and its Disruptor Variants
Base and disruptor variants of ElkarHizketak, built to stress-test conversational RAG systems in Basque under realistic interaction patterns (conversational openings, topic shifts).
Dataset Details
Dataset Description
This dataset extends ElkarHizketak with a base variant (rewritten opening queries, retrieval-needed labels, retrieved chunks) and disruptor variants that inject… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/elkarhizketak-RAG.Multilingual_POPEThis is the multilingual version of POPE. If you find it useful, we hope you can cite our paper.
@article{ye2025claim,
title={CLAIM: Mitigating Multilingual Object Hallucination in Large Vision-Language Models with Cross-Lingual Attention Intervention},
author={Ye, Zekai and Li, Qiming and Feng, Xiaocheng and Qin, Libo and Huang, Yichong and Li, Baohang and Jiang, Kui and Xiang, Yang and Zhang, Zhirui and Lu, Yunfei and others},
journal={arXiv preprint arXiv:2506.11073},
year={2025}
}
f-cov-offset1024-qwen3-1.7b-base-math12k-c50b0bba-rollouts
f_cov_Qwen3-1.7B-Base_math12k_offset1024_token_mean rollouts
This dataset contains one compressed JSONL shard for every completed training
step. The step and rollout_index columns uniquely locate a rollout within
this training run. Run metadata and per-step row counts are recorded in
rollout_manifest.json.
PAWS-eu
Dataset Card for PAWS-eu
Point of Contact: hitz@ehu.eus
Dataset Description
Dataset Summary
PAWS-eu is the professional translation to Basque of the PAWS dataset (Zhang et al., 2019),
in the spirit of the PAWS-X effort (Yang et al., 2019).
PAWS consist of sentence pairs that have high lexical overlap but that may or may not be paraphrases.
Languages
eu-ES
Dataset Structure
Data Fields
id (str): A unique id for each pair.… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/PAWS-eu.EusExams-v2
Dataset Card for EusExams-v2
EusExams-v2 is an updated and refined collection of tests designed to prepare individuals for Public Service examinations conducted by several Basque institutions, including the public health system Osakidetza, the Basque Government, the City Councils of Bilbao and Gasteiz, and the University of the Basque Country (UPV/EHU). Within each of these groups, there are different exams for public positions, such as administrative and assistant roles. Each… See the full description on the dataset page: https://huggingface.co/datasets/HiTZ/EusExams-v2.