AI2 ARC: Science Questions
7,887 multiple-choice science questions for evaluating question-answering models.
- Question answering
- English
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.
Answered by Gemini 4 · the chat is kept only in this tab
7,887 multiple-choice science questions for evaluating question-answering models.
1 to 10 thousand JSON records with scores of various models from leaderboards, research papers, and local runs.
10,000 short video clips with copyright status labels for copyright infringement detection systems.
Large-scale robot teleoperation recordings across 56 tasks using head and wrist cameras.
A set of 1,000 short video clips for training and testing automated copyright infringement search systems.
4.5 million time series and 230 billion data points for pre-training forecasting foundation models.
Massive corpus of annotated DNA nucleotide sequences for training bioinformatics and genomic models.
Collection of tasks and data for training and evaluating English text understanding systems.
100k to 1M synthetic images with combinations of concepts and ImageNet classes for evaluating multimodal models.
Large multimodal dataset for LLaVA-OneVision-1.5 intermediate training; source downloads completed.
1.3T tokens of filtered educational web pages in English for LLM training.
Large-scale multilingual archive of news articles from 2016–2026 with quality filtering and deduplication.
Continuation of the project slicing videos into audio clips with Whisper transcriptions for speech tasks.
Repository of scanned books and documents from the VOMEBOOK project for searching and working with text archives.
Large English web corpus broken down by dozens of domains and iteration statistics for training and classification.
Continuation of the speech recognition and synthesis dataset export gathered from video files.
Higgs boson mass measurement test dataset (H→ZZ*→4ℓ) intended strictly for model evaluation within the benchmark.
Documentation from all PRs updating Hugging Face Docs pages; automatically built via GitHub Action.
Large video dataset of robot interactions with the real environment for reinforcement learning.
Generated images from complex prompts to check composition and logical relations in text-to-image models.
Multimodal web interaction recordings formatted as dialogs for training web agents.
12k complex multi-discipline questions for rigorous LLM evaluation.
Collection of all published documentation from the Hugging Face website, generated using the doc-builder tool.
Massive dataset of 858,000 3D objects with detailed textures and English descriptions.