SFT examples for ultra-long LLM responses
6K English and Chinese SFT examples with responses of 2K to 32K words.
- Text generation
- English
- Chinese
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
6K English and Chinese SFT examples with responses of 2K to 32K words.
Consolidated corpus of user-AI assistant conversations for training chat models.
English texts on offensive security, red team, and pentesting for model training and evaluation.
English command-line action trajectories from NVIDIA for reinforcement learning of LLM agents.
SFT collection for mathematical and scientific reasoning, distilled from DeepSeek-V4-Pro.
10k to 100k English question-answer pairs gathered from the WikiDoc open medical encyclopedia.
An English dataset of 100K-1M examples with error breakdowns in responses for training models via critique.
Dataset of over 100,000 records for Indian Sign Language recognition and translation.
Multimodal English captions for image-text alignment and training captioning models.
Synthetic English instruction-response pairs for training agentic models.
Open corpus from NVIDIA of 1-10 million English profiles built on real British demographics and psychology.
English test suite for evaluating language models' knowledge in cybersecurity.
A Chinese text corpus from Zhihu for training Open Assistant, containing 1M to 10M examples.
Over 500,000 documents and 500 English questions for testing RAG systems on business data.
340 legal exams from Switzerland, the EU, and international law for evaluating reasoning models.
10k to 100k Korean question-answer pairs for training machine reading systems, built to the SQuAD 1.0 standard.
Over 100,000 records of AI agent shopping sessions for text-generation and RL tasks.
17,202 cleaned tasks for solving real Python bugs with validated patches.
157k synthetic discharge summaries with question-answer pairs for training clinical LLMs.
GSM8K-style math problems generated by Llama-3.1-405B with detailed thoughts and answers.
Over 150 million bibliographic records from the Open Library project for NLP, search, and data extraction tasks.
A dataset for training models on instruction-based text editing in English.
A massive corpus of chemical molecules in text representation for pre-training generative models.
Professionally translated Wikipedia sentences in 101 languages for evaluating multilingual translation quality.