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

Datasetnvidia/SPEED-Bench

NVIDIA LLM throughput benchmark

Long-prompt collection (up to 100K tokens) with varying entropy for measuring language model inference speed and quality.

    27.1K 541K–10K rows
    Datasetgaia-benchmark/GAIA

    GAIA Benchmark for Agentic LLMs

    450+ non-trivial questions with single-answer targets for testing tool-using and search-equipped LLMs.

    • English
    26.9K 999up to 1K rows🔒 on request
    DatasetRevolutionCrossroads/loc_chronicling_america_1770-1810

    OCR Archive of 1770-1810 American Newspapers

    Historical US newspaper pages with images, original OCR, and neural text recognition.

    • Image → text
    • Text retrieval
    • Text classification
    • English
    26.6K 810K–100K rows
    Datasetdatabricks/officeqa-pro-v2

    OfficeQA Pro v2 document reasoning benchmark

    Dataset for evaluating models and agents on US Federal report documents.

    • Question answering
    • Text generation
    • Text retrieval
    • English
    25.2K 25up to 1K rows🔒 on request
    DatasetOpenSourceAP/CrossSection

    CrossSection asset pricing signals

    A large-scale financial dataset with market signals and portfolio returns for quantitative analysis.

    • English
    24.3K 31M–10M rows
    DatasetIntelligenceLab/Long-Horizon-Terminal-Bench

    Long-Horizon Terminal Tasks Benchmark

    LHTB evaluates how LLM agents maintain useful work in a containerized terminal over hundreds of steps.

    • Text generation
    • English
    24.3K 144up to 1K rows
    Datasetpixparse/pdfa-eng-wds

    PDF pages for vision-language models

    English PDF pages from SafeDocs with annotations for training and evaluating OCR and vision-language models.

    • Image → text
    • English
    23.8K 16810M–100M rows
    DatasetRevolutionCrossroads/nara_revolutionary_war_pension_files_PDFs

    NARA Revolutionary War pension files archive

    10k to 100k digitized multi-page National Archives documents in English.

    • Image → text
    • Text retrieval
    • Text classification
    • English
    23.6K 010K–100K rows
    DatasetQiao123rvvr/eccv2026-cad-challenge-data

    ECCV 2026 CAD Challenge data

    Dataset for the 3D model generation from images competition within the ECCV 2026 workshop.

    • Image → 3d
    23.5K 01K–10K rows
    Datasettegridydev/research-papers

    Research Papers Collection in PDF and Markdown

    Small set of English research papers categorized by topic for OCR and document classification tests.

    • Image classification
    • Text classification
    • English
    22.9K 19up to 1K rows
    Datasethashmortar/spreadsheet-bench-v2-modified

    Table, PDF, and document question answering

    A set of 1,060 questions and reference answers in English across 171 business documents in various formats.

    • Question answering
    • Table question answering
    • Document question answering
    • English
    22.8K 21K–10K rows
    DatasetPaddlePaddle/Real5-OmniDocBench

    Distorted document parsing benchmark

    Document image set with physical distortions to test OCR and parsing system robustness.

    • Image → text
    • Chinese
    • English
    21.5K 431K–10K rows
    DatasetVincentHancoder/RAPTop pick

    Scientific Papers and Posters with HTML Markup

    25,435 triples of scientific papers, posters, and HTML/CSS templates from AI conferences for poster generation.

    • Image → text
    • Text → image
    • English
    20.8K 310K–100K rows
    Datasetllamaindex/ExtractBench

    Document data extraction benchmark

    Documents and schemas for testing structured JSON extraction with evidence.

    • English
    20.5K 74up to 1K rows
    Datasetheitorrosa/cvm-corpus

    Brazilian regulator CVM financial report corpus

    Cleaned and chunked documents from Brazilian regulator CVM in Portuguese for language model fine-tuning.

    • Text generation
    • Fill mask
    • Portuguese
    20K 0100K–1M rows
    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
    Datasetallenai/olmOCR-bench

    OCR bench for PDF to Markdown

    1,403 PDFs and 7,010 unit tests to evaluate OCR systems on converting PDFs to Markdown while preserving structure.

    • English
    17.9K 3351K–10K rows
    Datasetpixparse/idl-wds

    UCSF documents with OCR annotations

    Millions of document pages with PDF, TIFF images, and OCR annotations for recognition training.

    • Image → text
    16.9K 20710M–100M rows
    DatasetPeacefulData/RMS-AQA

    Spatial audio and household scene questions

    English benchmark for evaluating spatial sound understanding and multi-hop reasoning.

    • Question answering
    • English
    10.1K 15🔒 on request
    Dataseteve-bio/drug-target-activity

    Drug-target interaction activity

    Tabular and text data from laboratory activity screening of FDA-approved drugs across various targets.

    • English
    9.6K 43100K–1M rows🔒 on request
    Datasetgoogle/multimodal-knowledge-agentTop pick

    Google DeepMind benchmark for multimodal agents

    10 demonstration tasks from Google DeepMind for evaluating AI agents working with complex graphics and documents.

    • Image text → text
    • Visual question answering
    • Document question answering
    • English
    8.8K 103up to 1K rows
    Datasetmodusaudit/FinancialAuditBenchTop pick

    Financial statement audit benchmark

    A set of tasks to evaluate AI agents in performing financial statement audits.

    • English
    7.5K 5up to 1K rows
    Datasetdatalab-to/omni_parse_bench

    Document and table parsing benchmark

    A set of 16,288 tests across 2,937 pages for evaluating the accuracy of text, table, and layout recognition.

    • Image → text
    7.1K 22100K–1M rows
    Datasetulamai/UnsolvedMathTop pick

    Open math problems for reasoning benchmark

    17.6K published math problems of varying difficulty for testing reasoning models.

    • Question answering
    • Text generation
    • English
    6.9K 7910K–100K rows