Synthetic question-answer pairs for SQuADShifts
Between 1 and 10 million synthetic English question-answer pairs for evaluating question generation 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
Between 1 and 10 million synthetic English question-answer pairs for evaluating question generation models.
Multimodal dataset from Amazon Science to test VLM capability to find answers in bulky documents and images.
English and Chinese documents and tables for testing slide generation.
250 English test tasks to evaluate AI agents that study software and build replicas.
A corpus of Claude Code proxy server logs preserving timestamps and a 256k limit per request.
Mathematical reasoning benchmark over images: IQTest, FunctionQA, and PaperQA; languages en, zh, fa.
103K complex math problems with verifiable answers, decontaminated, for RL training.
Collection of open scientific texts and metadata from arXiv, CVF, NeurIPS, and other venues.
A massive 9.3-billion-row tabular corpus with complete financial statements of fictional firms over 31 years.
A massive corpus of scientific paper summaries with OpenAlex metadata and full texts of 5.3 million articles.
A multilingual collection of Wikimedia vector diagrams and maps with text coordinate annotations for evaluating image translation.
Reading comprehension and QA dataset featuring 100K SQuAD 1.1 questions plus 50K unanswerable questions on Wikipedia articles.
A manually corrected and verified version of the SciCode benchmark across 64 scientific programming tasks.
Over 1.3M PII-free posts from 110 news, IT, and official Telegram channels for language model training.
Medical research questions with yes/no/maybe answers based on PubMed abstracts.
English dataset of 60 thousand verified examples for training LLMs in tool and API use.
162 English tasks for evaluating LLM legal reasoning across classification and question answering.
English QA benchmark: real user questions requiring the reading of a Wikipedia article to answer.
Scientific coding task dataset authored by researchers to evaluate LLMs.
Multidisciplinary visual and textual questions for multimodal understanding evaluation.
1-10 billion English and Chinese texts after a quality check pipeline.
Over 256 thousand publicly shared chats with language models, updated weekly.
Documentation and code pairs from GitHub for natural language code search and code model training.
GitHub code with L0–L3 cleaning levels for training language models.