datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
jigsaw-toxic-comment-classification-challenge
Dataset Description
You are provided with a large number of Wikipedia comments which have been labeled by human raters for toxic behavior. The types of toxicity are:
toxic
severe_toxic
obscene
threat
insult
identity_hate
You must create a model which predicts a probability of each type of toxicity for each comment.
File descriptions
train.csv - the training set, contains comments with their binary labels
test.csv - the test set, you must predict the toxicity… See the full description on the dataset page: https://huggingface.co/datasets/thesofakillers/jigsaw-toxic-comment-classification-challenge.midi-classical-music
MIDI Classical Music
This dataset contains a comprehensive collection of MIDI files representing classical music compositions from various renowned composers.
The collection includes works from composers such as Bach, Beethoven, Chopin, Mozart, and many others.
The dataset is organized into directories by composer, with each directory containing MIDI files of their compositions.
The dataset is ideal for music analysis, machine learning models for music generation, and other… See the full description on the dataset page: https://huggingface.co/datasets/drengskapur/midi-classical-music.ec_classificationaidovecl-vehicle-detection-classification-localization
AIDOVECL: AI-generated Dataset of Outpainted Vehicles for Eye-level Classification and Localization
We introduce an annotated AI-generated dataset of eye-level vehicle images using outpainting, offering versatile generation of diverse vehicle classes in varied contexts with pretrained models.
Citation Notice
Please ensure that all publications and presentations using this data reference the following paper:
Kazemi, A., Fatima, Q. ul A., Kindratenko, V., & Tessum, C. W.… See the full description on the dataset page: https://huggingface.co/datasets/amir-kazemi/aidovecl-vehicle-detection-classification-localization.eai-taxonomy-math-w-fm-classify-behaviors
🧮 EAI Taxonomy Math w/ Behavioral Classifications (10K Sample)
A 10,000 document sample from EssentialAI/eai-taxonomy-math-w-fm enhanced with 4 behavioral reasoning classifications using GPT-4.1-mini.
Behavioral Classifications
Structured behavioral analysis following the approach from cognitive-behaviors:
backtracking_json: Identifies reasoning that backtracks or revisits earlier steps
backward_chaining_json: Detects goal-oriented reasoning working backwards… See the full description on the dataset page: https://huggingface.co/datasets/nlile/eai-taxonomy-math-w-fm-classify-behaviors.jailbreak-classification
Jailbreak Classification
Dataset Summary
Dataset used to classify prompts as jailbreak vs. benign.
Dataset Structure
Data Fields
prompt: an LLM prompt
type: classification label, either jailbreak or benign
Dataset Creation
Curation Rationale
Created to help detect & prevent harmful jailbreak prompts when users interact with LLMs.
Source Data
Jailbreak prompts sourced from: https://github.com/verazuo/jailbreak_llms… See the full description on the dataset page: https://huggingface.co/datasets/jackhhao/jailbreak-classification.classnodaikirainajoshitokekkonsurukotoninatta
Bangumi Image Base of Class No Daikirai Na Joshi To Kekkon Suru Koto Ni Natta.
This is the image base of bangumi Class no Daikirai na Joshi to Kekkon suru Koto ni Natta., we detected 33 characters, 3154 images in total. The full dataset is here.
Please note that these image bases are not guaranteed to be 100% cleaned, they may be noisy actual. If you intend to manually train models using this dataset, we recommend performing necessary preprocessing on the downloaded dataset to… See the full description on the dataset page: https://huggingface.co/datasets/BangumiBase/classnodaikirainajoshitokekkonsurukotoninatta.ILUR-news-text-classification-corpus
News Texts Dataset
We release a dataset of over 12000 news articles from iLur.am, categorized into 7 classes: sport, politics, weather, economy, accidents, art, society. The articles are split into train (2242k tokens) and test sets (425k tokens).
For more details, refer to the paper.
ClassEval
Dataset Card for FudanSELab ClassEval
Dataset Summary
We manually build ClassEval of 100 class-level Python coding tasks, consists of 100 classes and 412 methods, and average 33.1 test cases per class.
For 100 class-level tasks, diversity is maintained by encompassing these tasks over a wide spectrum of topics, including Management Systems, Data Formatting, Mathematical Operations, Game Development, File Handing, Database Operations and Natural Language Processing.
For… See the full description on the dataset page: https://huggingface.co/datasets/FudanSELab/ClassEval.cpc-classificationsclassimgman2000classimgman2000-2cybersecurity-classification-benchmark
TorchSight Cybersecurity Classification Benchmark
A two-tier benchmark dataset for evaluating cybersecurity document
classifiers, released with the TorchSight system. Used in:
Dobrovolskyi, I. Security Document Classification with a Fine-Tuned Local
Large Language Model: Benchmark Data and an Open-Source System. Journal of
Information Security and Applications, 2026.
Canonical per-model numbers live in BENCHMARK_NUMBERS.md,
auto-generated from the per-prediction result JSONs… See the full description on the dataset page: https://huggingface.co/datasets/torchsight/cybersecurity-classification-benchmark.ParlaSpeech-RS
The Serbian Parliamentary Spoken Dataset ParlaSpeech-RS 1.0
The master dataset can be found at http://hdl.handle.net/11356/1834.
Notice: ParlaSpeech corpora are currently in the process of enrichment with new features. Follow our progress here: http://clarinsi.github.io/parlaspeech
The ParlaSpeech-RS dataset is built from the transcripts of parliamentary proceedings available in the Serbian part of the ParlaMint corpus (http://hdl.handle.net/11356/1859), and the parliamentary… See the full description on the dataset page: https://huggingface.co/datasets/classla/ParlaSpeech-RS.ade_corpus_v2_classification
ADE-Corpus-V2 Dataset: Adverse Drug Reaction Data.
This is a dataset for classification if a sentence is ADE-related (True) or not (False).
Train size: 17,637
Test size: 5,879
Source dataset
Paper
arxiv-classificationArxiv Classification: a classification of Arxiv Papers (11 classes).
This dataset is intended for long context classification (documents have all > 4k tokens). Copied from "Long Document Classification From Local Word Glimpses via Recurrent Attention Learning"
@ARTICLE{8675939,
author={He, Jun and Wang, Liqun and Liu, Liu and Feng, Jiao and Wu, Hao},
journal={IEEE Access},
title={Long Document Classification From Local Word Glimpses via Recurrent Attention Learning},
year={2019}… See the full description on the dataset page: https://huggingface.co/datasets/ccdv/arxiv-classification.commit-classification-dataset
Commit Classification Dataset
This dataset is designed for multi-label classification of Git commit messages into predefined categories.
Dataset Summary
This dataset contains:
Training data: Commit messages and their corresponding labels for training the model.
Validation data: A separate set of messages for tuning and evaluation.
Testing data: Unlabeled commit messages for testing the model’s performance.
The goal of the dataset is to classify each commit message into… See the full description on the dataset page: https://huggingface.co/datasets/meriemm6/commit-classification-dataset.task903_deceptive_opinion_spam_classification
Dataset Card for Natural Instructions (https://github.com/allenai/natural-instructions) Task: task903_deceptive_opinion_spam_classification
Additional Information
Citation Information
The following paper introduces the corpus in detail. If you use the corpus in published work, please cite it:
@misc{wang2022supernaturalinstructionsgeneralizationdeclarativeinstructions,
title={Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+… See the full description on the dataset page: https://huggingface.co/datasets/Lots-of-LoRAs/task903_deceptive_opinion_spam_classification.molmo2-tulu4-classifiedthe-vault-classThe Vault is a multilingual code-text dataset with over 40 million pairs covering 10 popular programming languages.
It is the largest corpus containing parallel code-text data. By building upon The Stack, a massive raw code sample collection,
the Vault offers a comprehensive and clean resource for advancing research in code understanding and generation. It provides a
high-quality dataset that includes code-text pairs at multiple levels, such as class and inline-level, in addition to the function level.
The Vault can serve many purposes at multiple levels.task902_deceptive_opinion_spam_classification
Dataset Card for Natural Instructions (https://github.com/allenai/natural-instructions) Task: task902_deceptive_opinion_spam_classification
Additional Information
Citation Information
The following paper introduces the corpus in detail. If you use the corpus in published work, please cite it:
@misc{wang2022supernaturalinstructionsgeneralizationdeclarativeinstructions,
title={Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+… See the full description on the dataset page: https://huggingface.co/datasets/Lots-of-LoRAs/task902_deceptive_opinion_spam_classification.curio-rewrite-non-edu-dataset
Curio Rewrite — Non-Educational
Portuguese web texts (non-educational subset, sampled from ClassiCC) rewritten by Qwen2.5-7B-Instruct under four prompt styles. Companion to the educational subset; used to train the Curio rewrite models.
Config
Prompt style
Rows
easy
Simple vocabulary, child-friendly paraphrase
22,237,886
medium
Moderate paraphrase
14,698,285
hard
Sophisticated paraphrase
18,576,570
qa
Reformatted as question/answer
18,664,285
Fields… See the full description on the dataset page: https://huggingface.co/datasets/ClassiCC-Corpus/curio-rewrite-non-edu-dataset.ParlaSpeech-CZ
Dataset Card for "ParlaSpeech-CZ.v1.0"
The master dataset can be found at http://hdl.handle.net/11356/1785.
Notice: ParlaSpeech corpora are currently in the process of enrichment with new features. Follow our progress here: http://clarinsi.github.io/parlaspeech
The ParlaSpeech-CZ dataset is built from the transcripts of parliamentary proceedings available in the Czech part of the ParlaMint corpus (http://hdl.handle.net/11356/1859), and the parliamentary recordings available… See the full description on the dataset page: https://huggingface.co/datasets/classla/ParlaSpeech-CZ.classroomoftheelite
Bangumi Image Base of Classroom Of The Elite
This is the image base of bangumi Classroom of the Elite, we detected 58 characters, 4577 images in total. The full dataset is here.
Please note that these image bases are not guaranteed to be 100% cleaned, they may be noisy actual. If you intend to manually train models using this dataset, we recommend performing necessary preprocessing on the downloaded dataset to eliminate potential noisy samples (approximately 1% probability).… See the full description on the dataset page: https://huggingface.co/datasets/BangumiBase/classroomoftheelite.commit-classification-17kParlaSpeech-HR
The Croatian Parliamentary Spoken Dataset ParlaSpeech-HR 2.0
The master dataset can be found at http://hdl.handle.net/11356/1914.
Notice: ParlaSpeech corpora are currently in the process of enrichment with new features. Follow our progress here: http://clarinsi.github.io/parlaspeech
The ParlaSpeech-HR dataset is built from the transcripts of parliamentary proceedings available in the Croatian part of the ParlaMint corpus (http://hdl.handle.net/11356/1859), and the parliamentary… See the full description on the dataset page: https://huggingface.co/datasets/classla/ParlaSpeech-HR.ai-detector-data
AI Detector Predictions Dataset
A continuously-growing collection of AI text detection predictions with optional user feedback, generated from the AI Text Detector Space.
Every time someone analyzes text or a URL on the Space, the prediction is appended to this dataset. Users can also click "Correct" or "Incorrect" to provide feedback, which gets stored alongside the prediction.
Schema
Field
Type
Description
id
string
Unique 12-char hex identifier… See the full description on the dataset page: https://huggingface.co/datasets/adaptive-classifier/ai-detector-data.AfriMCQA-category-classification
Afri-MCQA cross-modal cultural category classification (MTEB)
Classify the cultural category of an entry from its photograph and the question
about it spoken by a native speaker, across 16 African languages.
Labels index this list:
geography, building, and landmarks
public figure and pop culture
cooking and food
objects, materials, clothing
tranditions, art, and history
brands, products, and companies
plants and animals
people, and everyday life
vehicles and transportation… See the full description on the dataset page: https://huggingface.co/datasets/vnahata/AfriMCQA-category-classification.ProfNER_corpus_classificationAraMix-domain-classified
AraMix Domain-Classified
AraMix family: AraMix (minhash and matched) | AraMix-domain-classified (with domain labels) | AraMix-HQ (model-filtered)
This is AraMix with per-document domain labels from nvidia/multilingual-domain-classifier.
Usage
from datasets import load_dataset
ds = load_dataset("AdaMLLab/AraMix-domain-classified", "minhash_deduped")
ds = load_dataset("AdaMLLab/AraMix-domain-classified", "sentence_deduped")
Schema
Field… See the full description on the dataset page: https://huggingface.co/datasets/AdaMLLab/AraMix-domain-classified.