Corpus and embeddings for agent search
Over 50 million documents and 1 billion passages with jina-v2-base embeddings for training search.
- Text generation
- 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
Over 50 million documents and 1 billion passages with jina-v2-base embeddings for training search.
Google Code project snapshot from 2006-2016 for training models on historical code.
A large-scale text dataset with difficulty grading for SFT and enhancing model reasoning skills.
A large multimodal set of 1 million to 10 million image-text pairs in English.
1.79M medical examples with reasoning from 7 models for SFT.
About 5 million JavaScript code files for pre-training and fine-tuning code models.
Interactive clinical benchmark for evaluating AI agents in metabolic medicine and GLP-1 therapy.
Pre-training corpus for generative models featuring math, code, and multilingual Q&A.
1 million to 10 million English examples with chains of thought for generation and classification.
Comprehensive test set of medical questions and answers for evaluating language models in Chinese.
English-language benchmark with logic puzzles for evaluating language model reasoning skills.
Around 300 thousand CVE vulnerability records formatted as multi-turn dialogues for AI agent training.
About 20,000 question-answer pairs with span-based answers for Chinese reading comprehension.
Metadata of 2.99 million scientific publications from 1991 to 2026 in parquet format for search and analytics.
Finance instruction-response pairs based on Alpaca and FiQA plus 1.3k synthetic examples.
99,870 system-user-assistant triplets for defensive cybersecurity SFT in English.
173 million eukaryotic and prokaryotic DNA and RNA sequences for pretraining base genomic models.
26,750 open prompts in over 100 languages for evaluating text generation quality.
Over 700,000 synthetic Chinese examples based on SmolTalk for SFT.
Nearly deduplicated version of The Stack v2—a large code corpus for LLM training, grouped by repositories.
1 to 10 million descriptions of virtual German residents generated based on real demographics.
Dataset for evaluating medical language models across 35 categories.
About 50k instruction-response pairs in Chinese generated by GPT-4 using the Alpaca method.
Over 33,000 real patient questions and 127,000 doctor answers in Portuguese.