Common sense benchmark for 136 language variants
Multiple-choice questions to test understanding of everyday logic and cultural features in rare languages.
- Question answering
- Multiple choice
- acm
- acq
- Tunisian Arabic
- +9
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
Multiple-choice questions to test understanding of everyday logic and cultural features in rare languages.
10 to 100 thousand English examples for training models to work with hidden user profiles and memory.
Cleaned English corpus of 400 billion tokens for efficient language model pre-training.
Large Chinese pre-training corpus; version V2 is marked as deprecated.
3.3k expert pairwise preference annotations for six model responses to 80 MT-bench questions.
Data for training and evaluating safe responses: preferences, question-answer pairs, and English prompts.
Code in many programming languages with permissive licenses after near-deduplication, about 1.5 TB.
Dataset for research on extracting hidden reasoning chains from proprietary APIs on complex tasks.
35,030 expert multiple-choice questions across 45 subjects in Korean.
385K human preferences for answer helpfulness across 18 topics.
A large corpus of StackOverflow questions, answers, and code up to June 14, 2023, in Markdown text.
A large multilingual collection of texts for classification, summarization, and translation.
Millions of English and Chinese instructions for LLM fine-tuning and evaluation.
Updated dataset based on new human evaluations for testing reward models (RLHF).
A synthetic corpus of simple stories for training small and interpretable language models.
Cleaned preference pairs from UltraFeedback for language model fine-tuning.
4,006 raw agent traces with DeepSeek v4 Pro responses and a tool schema snapshot for SFT.
Large English FineWeb-Edu corpus after precise deduplication, featuring embeddings and row repetition counters.
Compact dataset of fewer than 1000 hazardous English prompts for evaluating language model guardrails.
Text and tabular tasks for training and evaluating Chinese language understanding models.
Binarized preferences for DPO based on Capybara: multi-turn dialogues for training chat models.
Filtered code corpus of ~240B tokens across 11 programming languages from The Stack v3.
An English dataset for LLM pretraining focused on high quality and efficient scaling.
2.7 million US wire service articles from 1878-1977 with geolocation, topics, and entities.