datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
FineFineWeb
FineFineWeb: A Comprehensive Study on Fine-Grained Domain Web Corpus
arXiv: Coming Soon
Project Page: Coming Soon
Blog: Coming Soon
Data Statistics
Domain (#tokens/#samples)
Iteration 1 Tokens
Iteration 2 Tokens
Iteration 3 Tokens
Total Tokens
Iteration 1 Count
Iteration 2 Count
Iteration 3 Count
Total Count
aerospace
5.77B
261.63M
309.33M
6.34B
9100000
688505
611034
10399539
agronomy
13.08B
947.41M
229.04M
14.26B
15752828
2711790
649404
19114022… See the full description on the dataset page: https://huggingface.co/datasets/m-a-p/FineFineWeb.glue
Dataset Card for GLUE
Dataset Summary
GLUE, the General Language Understanding Evaluation benchmark (https://gluebenchmark.com/) is a collection of resources for training, evaluating, and analyzing natural language understanding systems.
Supported Tasks and Leaderboards
The leaderboard for the GLUE benchmark can be found at this address. It comprises the following tasks:
ax
A manually-curated evaluation dataset for fine-grained… See the full description on the dataset page: https://huggingface.co/datasets/nyu-mll/glue.super_glue
Dataset Card for "super_glue"
Dataset Summary
SuperGLUE (https://super.gluebenchmark.com/) is a new benchmark styled after
GLUE with a new set of more difficult language understanding tasks, improved
resources, and a new public leaderboard.
Supported Tasks and Leaderboards
More Information Needed
Languages
More Information Needed
Dataset Structure
Data Instances
axb
Size of downloaded dataset files: 0.03 MB
Size of… See the full description on the dataset page: https://huggingface.co/datasets/aps/super_glue.dclm-baseline-1.0
DCLM-baseline
DCLM-baseline is a 4T token / 3B document pretraining dataset that achieves strong performance on language model benchmarks.
Below are comparisions of model trained on DCLM-baseline with other models in the 7B regime.
Model
Params
Tokens
Open dataset?
CORE
MMLU
EXTENDED
Open weights, closed datasets
Llama2
7B
2T
✗
49.2
45.8
34.1
DeepSeek
7B
2T
✗
50.7
48.5
35.3
Mistral-0.3
7B
?
✗
57.0
62.7
45.1
QWEN-2
7B
?
✗
57.5
71.9
50.5
Llama3
8B
15T
✗… See the full description on the dataset page: https://huggingface.co/datasets/mlfoundations/dclm-baseline-1.0.standard-chess-games
[!CAUTION]
This dataset is still a work in progress and some breaking changes might occur.
Lichess Rated Standard Chess Games Dataset
Dataset Description
6,771,826,271 standard rated games, played on lichess.org, updated monthly from the database dumps.
This version of the data is meant for data analysis. If you need PGN files you can find those here. That said, once you have a subset of interest, it is trivial to convert it back to PGN as shown in the Dataset Usage… See the full description on the dataset page: https://huggingface.co/datasets/Lichess/standard-chess-games.openbookqa
Dataset Card for OpenBookQA
Dataset Summary
OpenBookQA aims to promote research in advanced question-answering, probing a deeper understanding of both the topic
(with salient facts summarized as an open book, also provided with the dataset) and the language it is expressed in. In
particular, it contains questions that require multi-step reasoning, use of additional common and commonsense knowledge,
and rich text comprehension.
OpenBookQA is a new kind of… See the full description on the dataset page: https://huggingface.co/datasets/allenai/openbookqa.FineWeb-HQ
FineWeb-HQ
Dataset Summary
FineWeb-HQ is a high-quality, model-filtered pretraining dataset derived as a subset of FineWeb. FineWeb-HQ was created by selecting the top 10% of FineWeb documents based on a deep learning classifier trained to identify structured and knowledge-rich samples. This classifier uses XLM-RoBERTa embeddings to score documents.
To validate our approach, we pretrained 1B-parameter LLM models with a Llama-like architecture across multiple… See the full description on the dataset page: https://huggingface.co/datasets/epfml/FineWeb-HQ.pretraining_v1-omega_booksfineweb-edu
📚 FineWeb-Edu
1.3 trillion tokens of the finest educational data the 🌐 web has to offer
Paper: https://arxiv.org/abs/2406.17557
What is it?
📚 FineWeb-Edu dataset consists of 1.3T tokens and 5.4T tokens (FineWeb-Edu-score-2) of educational web pages filtered from 🍷 FineWeb dataset. This is the 1.3 trillion version.
To enhance FineWeb's quality, we developed an educational quality classifier using annotations generated by LLama3-70B-Instruct. We… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/fineweb-edu.fineweb
🍷 FineWeb
15 trillion tokens of the finest data the 🌐 web has to offer
What is it?
The 🍷 FineWeb dataset consists of more than 18.5T tokens (originally 15T tokens) of cleaned and deduplicated english web data from CommonCrawl. The data processing pipeline is optimized for LLM performance and ran on the 🏭 datatrove library, our large scale data processing library.
🍷 FineWeb was originally meant to be a fully open replication of 🦅 RefinedWeb, with a… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/fineweb.finephrase
Dataset Card for HuggingFaceFW/finephrase
Dataset Summary
Synthetic data generated by DataTrove:
Model: HuggingFaceTB/SmolLM2-1.7B-Instruct (main)
Source dataset: HuggingFaceFW/fineweb-edu, config sample-350BT, split train
Generation config: temperature=1.0, top_p=1.0, top_k=50, max_tokens=2048, model_max_context=8192
Speculative decoding: {"method":"suffix","num_speculative_tokens":32}
System prompt: None
Input column: text
Prompt families:
faq prompt
Rewrite… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceFW/finephrase.Rosetta-Activations
Rosetta Activations
Updated: 2026-06-15 02:30 UTC
Contrastive activation extractions for 17 semantic concepts across 46 language models,
supporting cross-architecture mechanistic interpretability research.
Companion concept pair corpus: jamesrahenry/Rosetta_Concept_Pairs
Papers: forthcoming
Dataset Structure
Rosetta-Activations/
├── rcp_v1/ # Current extraction line — richest data (N≈2000)
│ └── {Model_Name}/
│ ├── calibration_{concept}.npy… See the full description on the dataset page: https://huggingface.co/datasets/james-ra-henry/Rosetta-Activations.MMLU-Pro
MMLU-Pro Dataset
MMLU-Pro dataset is a more robust and challenging massive multi-task understanding dataset tailored to more rigorously benchmark large language models' capabilities. This dataset contains 12K complex questions across various disciplines.
|Github | 🏆Leaderboard | 📖Paper |
🚀 What's New
[2026.03.11] Added more cutting-edge frontier models to the leaderboard, including the Claude-4.6 series, Seed2.0 series, Qwen3.5 series, and Gemini-3.1-Pro… See the full description on the dataset page: https://huggingface.co/datasets/TIGER-Lab/MMLU-Pro.RekaDaily-10k-raw
RekaDaily-10k (raw)
Raw, unscripted, first-person daily-life video, collected through
Claru, Reka's data collection marketplace — recorded by
paid collectors in their own homes and workplaces on head-mounted and handheld
phones, across multiple regions.
Videos are delivered as recorded — no cuts, no trimming, no editing, no
filtering beyond basic integrity checks. A processed tier (short clips with
machine captions) is released separately under the same RekaDaily-10k prefix.… See the full description on the dataset page: https://huggingface.co/datasets/RekaAI/RekaDaily-10k-raw.stack-v3-train
🥞 The Stack v3
What is it?
What is being released
How to download and use it
Dataset statistics
Dataset structure
Dataset creation
Considerations for using the data
Additional information
What is it?
The Stack v3 is the largest, most up-to-date open dataset of source code, crawled directly from GitHub and built to pre-train code LLMs with full-repository context. It is the successor to The Stack v2 and, like its predecessor, is released to make the training… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceCode/stack-v3-train.common_corpus
Common Corpus
Full paper - ICLR 2026 oral
Common Corpus is the largest open licensed text dataset, comprising 2.27 trillion tokens (2,267,302,720,836 tokens). It is a diverse dataset, consisting of books, newspapers, scientific articles, government and legal documents, code, and more. Common Corpus has been created by Pleias in association with several partners.
Common Corpus differs from existing open datasets in that it is:
Truly Open: contains only data that is either… See the full description on the dataset page: https://huggingface.co/datasets/PleIAs/common_corpus.iclr-wm-backup-public
ICLR Watermark Benchmark — backup overflow (public part)
Companion to the private repo Aak975/iclr-wm-backup, which reached its
storage quota. Together the two repos form ONE backup — every file exists in
exactly one of them, with the same layout:
archives/<sub>/part-0000 ... part-NNNN, MANIFEST.json
restore one archive: cat part-* | zstd -d | tar -x
MANIFEST.json = {"parts": N, "sha256": <whole-stream>, "total_bytes": M}
This public part holds only shareable image data… See the full description on the dataset page: https://huggingface.co/datasets/Aak975/iclr-wm-backup-public.HPLT2.0_cleanedNB: HPLT2.0 is now superseded by a newer release:
HPLT3.0
We recommed switching to v3.0, unless you have a compelling reason to stay on 2.0.
This is a large-scale collection of web-crawled documents in 191 world languages, produced by the HPLT project.
The source of the data is mostly Internet Archive with some additions from Common Crawl.
For a detailed description of the dataset, please refer to our website and our pre-print.
The Cleaned variant of HPLT Datasets v2.0
This is… See the full description on the dataset page: https://huggingface.co/datasets/HPLT/HPLT2.0_cleaned.HelpSteer2
HelpSteer2: Open-source dataset for training top-performing reward models
HelpSteer2 is an open-source Helpfulness Dataset (CC-BY-4.0) that supports aligning models to become more helpful, factually correct and coherent, while being adjustable in terms of the complexity and verbosity of its responses.
This dataset has been created in partnership with Scale AI.
When used to tune a Llama 3.1 70B Instruct Model, we achieve 94.1% on RewardBench, which makes it the best Reward Model as… See the full description on the dataset page: https://huggingface.co/datasets/nvidia/HelpSteer2.EmbodiedGenDatahttps://huggingface.co/spaces/HorizonRobotics/EmbodiedGen-Gallery-Explorer
AI-CUDA-Engineer-Archive
The AI CUDA Engineer Archive 👷: Agentic CUDA Kernel Discovery, Optimization & Composition
We release The AI CUDA Engineer archive, a dataset consisting of approximately 30,000 CUDA kernels generated by The AI CUDA Engineer. It is released under the CC-By-4.0 license and can be accessed via HuggingFace and interactively visualized here. The dataset is based on the Kernel tasks provided in KernelBench and includes a torch reference implementation, torch, NCU and Clang-tidy… See the full description on the dataset page: https://huggingface.co/datasets/SakanaAI/AI-CUDA-Engineer-Archive.waqfeya-library
Waqfeya Library
📖 Overview
Waqfeya is one of the primary online resources for Islamic books, similar to Shamela. It hosts more than 10,000 PDF books across over 80 categories.
In this dataset, we processed the original PDF files using Google Document AI APIs and extracted their contents into two additional formats: TXT and DOCX.
📊 Dataset Contents
The dataset includes 22,443 PDF files (spanning 8,978,634 pages) representing 10,150 Islamic books. Each book is… See the full description on the dataset page: https://huggingface.co/datasets/ieasybooks-org/waqfeya-library.gpqa
Dataset Card for GPQA
GPQA is a multiple-choice, Q&A dataset of very hard questions written and validated by experts in biology, physics, and chemistry. When attempting questions out of their own domain (e.g., a physicist answers a chemistry question), these experts get only 34% accuracy, despite spending >30m with full access to Google.
We request that you do not reveal examples from this dataset in plain text or images online, to reduce the risk of leakage into foundation… See the full description on the dataset page: https://huggingface.co/datasets/Idavidrein/gpqa.witHiFi-UMI-2K
HiFi-UMI-2K: High-Fidelity Robot-Free Manipulation Data
2,000 hours released · 6 synchronized camera views · 480+ scenes · 3 mm pose accuracy · <40 µs synchronization
🌐 Project Website |
📦 Dataset |
📄 Paper: arXiv:2607.25895
Examples from the HiFi-UMI corpus. Click the image to play the video.
📚 Introduction
HiFi-UMI is a portable, high-fidelity bimanual capture system for collecting robot-free manipulation demonstrations.… See the full description on the dataset page: https://huggingface.co/datasets/simple-world-lab/HiFi-UMI-2K.RealCam-Vid
RealCam-Vid Dataset
News
25/04/08: We provide torch dataset demo code for example usage of our RealCam-Vid.
25/03/26: Release our dataset RealCam-Vid v1 for metric-scale camera-controlled video generation, containing ~100K video clips with dedicated short/long captions and metric-scale camera annotations.
25/02/18: Initial commit of the project, we plan to release the full dataset and data processing code in several… See the full description on the dataset page: https://huggingface.co/datasets/MuteApo/RealCam-Vid.Scientific-Summaries
Scientific Summaries
22 million LLM-generated structured summaries of scientific papers, enriched with OpenAlex scholarly metadata. Each paper has an 18-field structured summary covering methodology, key results, claims, limitations, and more. This public dataset includes full paper text for ~5.3 million papers where open-access status has been confirmed -- either through OpenAlex metadata or because the paper originates from a permissively licensed source such as the arXiv preprint… See the full description on the dataset page: https://huggingface.co/datasets/laion/Scientific-Summaries.datacomp200m
Datacomp200m
This is a smaller version of the datacomp_1b dataset.
Filtering was done by taking all rows that had self similarity (inner product) above 0.32. This resulted in 213009083 (213 million) rows.
The results of the datacomp paper suggest that filtering by CLIP score is better than random sampling.
Included in this repo are search indices created using autofaiss, over the text and image embeddings. There are two ways to access metadata, either in .parquet files in the… See the full description on the dataset page: https://huggingface.co/datasets/adams-story/datacomp200m.headqaExperimentDATA_knowledge_distillation_vs_fine_tuning