Skip to content

Datasets

Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.

Data type

For which models

Find a dataset for your goal

Answered by Gemini 4 · the chat is kept only in this tab

DatasetEdinburghNLP/xsum

XSum: BBC news brief summaries

BBC news articles with single-sentence summaries for extreme summarization training.

  • Summarization
  • English
19.6K 158100K–1M rows
Datasetjupyter-agent/jupyter-agent-dataset

QA and reasoning on Kaggle notebooks

Question-answer pairs and executable reasoning traces on real Kaggle notebooks for data analysis.

  • Question answering
  • Text generation
  • Code
19.4K 18210K–100K rows
Datasetprinceton-nlp/CharXiv

Benchmark for analyzing complex scientific charts

2,323 scientific charts from arXiv papers to test multimodal model ability to understand complex visualizations.

  • Visual question answering
  • Document question answering
  • Question answering
  • English
19.2K 571K–10K rows
Datasetscholarweave/arxiv-latexTop pick

ArXiv LaTeX sources with metadata

The complete corpus of arXiv LaTeX sources, parsed and aligned with official metadata in Parquet format.

  • Text generation
  • Embeddings
  • English
18.9K 1501M–10M rows
Datasetuonlp/CulturaXTop pick

Multilingual corpus in 167 languages

Cleaned multilingual corpus in 167 languages with 6.3 trillion tokens for training LLMs.

  • Text generation
  • Fill mask
  • Afrikaans
  • Albanian
  • Amharic
  • +9
18.9K 7121B–10B rows🔒 on request
DatasetSALT-NLP/SWE-chat

Real-world AI coding sessions

Developer dialogues with AI assistants: tool calls, reasoning traces, and code edits.

  • Text generation
  • English
  • Japanese
  • Chinese
  • +3
18.3K 17510M–100M rows🔒 on request
DatasetArtificialAnalysis/AA-Omniscience-Public

LLM factuality and hallucination benchmark

600 questions across various knowledge domains to evaluate factual accuracy and LLM hallucination tendency.

  • Question answering
18.2K 57up to 1K rows
DatasetManusagents/GPT-5.5-Gemini-3.1-Pro-Grok-4-Claude-Fable-5-Mythos-5-Qwen-3.7-Max-and-more-Distillation-Dataset

Distillation dataset from 73 models

18M+ distillation signals from 73 open models and 7k GitHub repositories for training text models.

  • Text generation
  • Other
  • English
  • Multilingual
18.2K 29010M–100M rows
Datasetopenbmb/UltraFeedback

UltraFeedback preference dataset

64K prompts and 256K LLM responses with ratings for training reward and critic models.

  • Text generation
  • English
17.8K 44910K–100K rows
Datasetmrlbenchmarks/global-piqa-nonparallelTop pick

Common sense benchmark for 136 language variants

Multiple-choice questions to test understanding of everyday logic and cultural features in rare languages.

  • Question answering
  • Multiple choice
  • acm
  • acq
  • Tunisian Arabic
  • +9
17.6K 4310K–100K rows
Datasettasksource/tasksource-jev-typed-decisions

Typed decisions from hundreds of Tasksource tasks

Millions of classification, choice, rating, and visual reasoning examples for multi-task learning.

  • Zero shot classification
  • Text classification
  • Question answering
  • English
  • Multilingual
17.5K 341M–10M rows
Datasetbowen-upenn/PersonaMem-v2

Dataset for LLM personalization and agent memory

10 to 100 thousand English examples for training models to work with hidden user profiles and memory.

  • Question answering
  • Text generation
  • English
17.3K 4010K–100K rows
Datasetnvidia/Nemotron-Math-v2

Nemotron-Math-v2: mathematical reasoning

347K math problems and 7M model-generated reasoning trajectories in English.

  • Text generation
  • English
17.2K 2041M–10M rows
Datasetnewfacade/LeetCodeDataset

Python LeetCode Dataset for LLMs

Python LeetCode problems for training and evaluating code LLMs; English text, 1k-10k examples.

  • Text generation
  • English
17.2K 871K–10K rows
DatasetSalesforce/xlam-function-calling-60k

xLAM: 60k function calling examples

Verified English function-calling examples for training and evaluating LLM agents.

  • Question answering
  • Text generation
  • Reinforcement learning
  • English
17K 76610K–100K rows🔒 on request
Datasetallenai/olmo-mix-1124Top pick

OLMo 2 pre-training set (November 2024)

Pre-training corpus for OLMo-2-1124: DCLM-Baseline, Arxiv, pes2o, StarCoder, and others.

  • Text generation
  • English
16.6K 931B–10B rows
DatasetTeraflopAI/SEC-EDGARTop pick

43 billion tokens of financial reports from SEC EDGAR

A massive corpus of company documents and reports from SEC EDGAR for training and evaluating financial language models.

  • Text generation
  • Text classification
  • English
16.4K 521M–10M rows
Datasetgoogle/RSRCCTop pick

Google Remote Sensing Change Captioning Benchmark

A Google Research dataset with multi-temporal aerospace image pairs and questions about terrain changes.

  • Visual question answering
  • Image text → text
  • Multiple choice
  • English
16.2K 41100K–1M rows
DatasetMegaScience/MegaScienceTop pick

Scientific reasoning for post-training

1.25 million open-source text examples for training and evaluating scientific reasoning.

  • Text generation
  • English
16.2K 1411M–10M rows
Datasetccore/ThreeJS-3D-ScenesTop pick

Three.js 3D scenes with prompts and code

4,146 examples of Three.js 3D scenes with prompts, thinking traces, and ready-to-use HTML/JavaScript code.

  • Text generation
  • English
16.1K 1821K–10K rows