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

Datasetlmqg/qa_squadshifts_synthetic

Synthetic question-answer pairs for SQuADShifts

Between 1 and 10 million synthetic English question-answer pairs for evaluating question generation models.

  • Question answering
  • English
47.8K 41M–10M rows
DatasetAmazonScience/document-haystack

Long Document Analysis Benchmark for Visual LLMs

Multimodal dataset from Amazon Science to test VLM capability to find answers in bulky documents and images.

  • Question answering
  • Visual question answering
  • Document question answering
  • English
47.6K 26
DatasetPresentBench/PresentBench

Presentation generation evaluation benchmark

English and Chinese documents and tables for testing slide generation.

  • Multimodal
  • Text generation
  • English
  • Chinese
47K 16up to 1K rows
DatasetQwen/RecreationBenchTop pick

Benchmark for app recreation agents

250 English test tasks to evaluate AI agents that study software and build replicas.

  • Text generation
  • English
46.7K 19up to 1K rows
Datasetsemianalysisai/cc-traces-weka-062126-256k

SemiAnalysis Claude Code request traces

A corpus of Claude Code proxy server logs preserving timestamps and a 256k limit per request.

  • Text generation
46.6K 8up to 1K rows
DatasetAI4Math/MathVista

MathVista: Visual-mathematical benchmark

Mathematical reasoning benchmark over images: IQTest, FunctionQA, and PaperQA; languages en, zh, fa.

  • Multiple choice
  • Question answering
  • Visual question answering
  • English
  • Chinese
  • Persian
45.9K 2371K–10K rows
Datasetzwhe99/DeepMath-103K

Complex Verifiable Math Problems for RL

103K complex math problems with verifiable answers, decontaminated, for RL training.

  • Text generation
  • English
45.6K 394100K–1M rows
DatasetJonesLin/writing-model-papers-2016-2021

Corpus of scientific papers 2016-2021

Collection of open scientific texts and metadata from arXiv, CVF, NeurIPS, and other venues.

  • Text generation
  • English
45.5K 010K–100K rows
Datasetyhcelebi/enterprise-dataTop pick

Synthetic Accounting and Operations of 30 Companies

A massive 9.3-billion-row tabular corpus with complete financial statements of fictional firms over 31 years.

  • Tabular classification
  • Tabular regression
  • Time series forecasting
  • English
  • Turkish
45.2K 21B–10B rows
Datasetlaion/Scientific-SummariesTop pick

22 million structured scientific article summaries

A massive corpus of scientific paper summaries with OpenAlex metadata and full texts of 5.3 million articles.

  • Summarization
  • Text classification
  • Embeddings
  • English
  • Multilingual
45.2K 3510M–100M rows
Datasetsarkau/ImgMT-Dataset-For-Evaluating-Text-Image-TranslationTop pick

Infographic and Diagram Text Translation Benchmark

A multilingual collection of Wikimedia vector diagrams and maps with text coordinate annotations for evaluating image translation.

  • Image text → text
  • Image text → image
  • Translation
44K 010K–100K rows
Datasetrajpurkar/squad_v2

Question answering on Wikipedia articles

Reading comprehension and QA dataset featuring 100K SQuAD 1.1 questions plus 50K unanswerable questions on Wikipedia articles.

  • Question answering
  • English
43.9K 275100K–1M rows
Datasetshhu2001/SciCode-Verified

Verified Scientific Code Generation Benchmark

A manually corrected and verified version of the SciCode benchmark across 64 scientific programming tasks.

  • Text generation
  • English
43.7K 2up to 1K rows
Datasethausmer/ua-ru-tg-channels-cleanTop pick

Cleaned Telegram channel posts in Russian and Ukrainian

Over 1.3M PII-free posts from 110 news, IT, and official Telegram channels for language model training.

  • Text generation
  • Ukrainian
  • Russian
43K 01M–10M rows🔒 on request
Datasetqiaojin/PubMedQA

Medical Yes/No Question Answering Dataset

Medical research questions with yes/no/maybe answers based on PubMed abstracts.

  • Question answering
  • English
41.9K 360100K–1M rows
Datasetlockon/xlam-function-calling-60k

60K Verified Function Calling Examples

English dataset of 60 thousand verified examples for training LLMs in tool and API use.

  • Question answering
  • Text generation
  • Reinforcement learning
  • English
41.7K 110K–100K rows
Datasetnguha/legalbench

LegalBench legal reasoning benchmark

162 English tasks for evaluating LLM legal reasoning across classification and question answering.

  • Text classification
  • Question answering
  • English
41.4K 20010K–100K rows
Datasetgoogle-research-datasets/natural_questions

User questions based on Wikipedia articles

English QA benchmark: real user questions requiring the reading of a Wikipedia article to answer.

  • Question answering
  • English
40.9K 13310K–100K rows
DatasetMMMU/MMMU_Pro

MMMU-Pro multimodal benchmark

Multidisciplinary visual and textual questions for multimodal understanding evaluation.

  • Question answering
  • Visual question answering
  • Multiple choice
  • English
40.4K 1031K–10K rows
Datasetopenbmb/Ultra-FineWebTop pick

Large Filtered En/Zh Web Corpus

1-10 billion English and Chinese texts after a quality check pipeline.

  • Text generation
  • English
  • Chinese
40.1K 4661B–10B rows
DatasetDJRHails/sharetrawl

Fresh public LLM dialogues from 9 platforms

Over 256 thousand publicly shared chats with language models, updated weekly.

  • Text generation
  • Question answering
  • English
  • Multilingual
39.7K 3100K–1M rows🔒 on request
Datasetcode-search-net/code_search_net

2M comment-code pairs

Documentation and code pairs from GitHub for natural language code search and code model training.

  • Text generation
  • Fill mask
  • Code
39.6K 3541M–10M rows
Datasetopenbmb/UltraData-Code

GitHub Code with L0–L3 Cleaning Levels

GitHub code with L0–L3 cleaning levels for training language models.

  • Text generation
  • English
  • Chinese
39.4K 216100M–1B rows