SuperGLUE language understanding benchmark
Challenging NLU tasks including textual entailment, QA, and coreference resolution for language model evaluation.
- Text classification
- Token classification
- 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
Challenging NLU tasks including textual entailment, QA, and coreference resolution for language model evaluation.
An open multilingual corpus of 2.27 trillion tokens spanning books, press, science, law, and code.
44k fill-in-the-blank tasks with two answer choices for evaluating commonsense reasoning.
Around 500 verifiable instructions for evaluating how well LLMs follow explicit constraints.
Unedited daily and working first-person video recordings in English for training video models.
Over 1 million trajectories from 100 robots, spanning 2976 hours and 100+ real-world scenarios.
500 human-verified issue-pull request pairs from Python repositories for evaluating automated GitHub issue resolution.
Over 36.7 million FineWeb-Edu documents automatically translated into 36 languages for training multilingual LLMs and translation systems.
Around 100k–1M 1080p videos and text descriptions for training video generation models.
Multimodal corpus of parametric 3D models from Objaverse featuring sand3d code, OpenUSD scenes, and images.
English synthetic text data generated by SmolLM2 from the FineWeb-Edu corpus.
Corpus of over 4.6 thousand Arabic books (over 5 million pages) with extracted text.
10k+ math problems with solutions for training and evaluating reasoning models.
1,140 tasks to evaluate models on prompt-based code generation and docstring-based completion.
A dataset for training and evaluating AI agents on Excel spreadsheet editing using reinforcement learning.
Multimodal dataset with video, 3D hand and object poses in real-world conditions for CVPR 2026.
The largest open collection of 10,000+ .pptx files with open licenses for slide analysis and generation.
Research corpus of 37.7M tokens for training EAGLE-style speculative decoders.
English QA benchmark with questions requiring multi-step reasoning and general knowledge.
50k labeled movie reviews for binary sentiment classification in English.
Reading comprehension QA dataset: questions on Wikipedia articles with span answers.
Spoken Egyptian Arabic dialect audio corpus with transcripts for TTS and ASR training.
Large text benchmark DABstep by Adyen for evaluating AI agents on multi-step reasoning tasks.
Over 70k free books in multiple languages for text generation.