Multilingual culturally-aware VQA benchmark
Over 10,000 multilingual image-question pairs with cultural context for evaluating multimodal models.
- Question answering
- Indonesian
- Sundanese
- Japanese
- +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
Over 10,000 multilingual image-question pairs with cultural context for evaluating multimodal models.
Large cleaned Uzbek language corpus in Latin and Cyrillic scripts for model pre-training.
English prompt-response pairs from Inkling and Qwen 3.8-27B models with teacher review for alignment.
A small English text dataset for question-answering tasks and sentence semantic similarity comparison.
A small set of ChatGPT jailbreak prompts for studying constraint bypassing.
1 to 10 million synthetic Indian resident profiles in English and Hindi for data generation and role-playing models.
A selection of real, complex English prompts from WildChat for evaluating large language models.
Cleaned reasoning dataset for GLM-5.1 with 1M examples, formatted to a unified SFT schema.
Hundreds of thousands of texts in six European languages for training models to hide personal information.
Comprehensive dataset of 180,000 question-answer pairs in 9 languages for end-to-end QA systems.
10,000 manually verified instruction-response pairs in Arabic for SFT.
3.4 million instruction-response pairs across 52 languages, translated from English alpaca-52k and dolly-15k.
An English question-answering dataset for code generation systems with improvement.
About 6 million question-answer pairs in 21 languages collected from the internet for search and QA systems.
7,469 test questions from South Korean medical licensing exams.
A collection of 21,318 prompts and attacks (jailbreaks, toxicity) for testing and improving neural network safety.
860 public English examples to evaluate model response fact-grounding.
7,716 JSONL question-answer examples for SFT distillation of reasoning models.
Preprocessed corpus of German court decisions for training and fine-tuning language models.
Text QA examples with synthetic reasoning from Claude, without human verification.
About 4.5 million synthetic FLANv2 examples with GPT-4 and GPT-3.5 answers for instruction tuning.
Curated 3D scenes with prompts, reasoning, and HTML/JavaScript code for generating 3D graphics.
English dataset of NeurIPS 2026 papers with key claims for reproduction.
Millions of question-answer pairs with document collections and pre-computed Qwen3-Embedding-4B vectors.