Multi-Teacher Distillation for SFT
57,937 filtered trajectories from three models for SFT in math, code, reasoning, and multilingual tasks.
- Text generation
- English
- Chinese
- Spanish
- +3
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
57,937 filtered trajectories from three models for SFT in math, code, reasoning, and multilingual tasks.
Dataset for supervised fine-tuning in instruction following and logical reasoning.
Parallel texts for evaluating translation from English into multiple languages, including low-resource ones.
Generated texts from various domains with PII and PHI entity annotations for training anonymization models.
Filtered reasoning traces from DeepSeek, Qwen, Gemma, and other models for small model distillation.
Multilingual real user dialogues with ChatGPT for chat model training and analysis.
8 Korean NLU tasks for evaluating language models: from NER and NLI to MRC and DST.
A set of adversarial English prompts for testing LLMs for safety policy compliance.
134 real-world tasks where agents learn in executable environments via multi-level feedback over 12+ hours.
708k English news articles from 2017–2019 sourced from Common Crawl.
A small collection of JSONL traces of AI agent operation with saved tool schemas for SFT.
About 70k problems with reasoning traces and code execution for training TIR models.
Texts from the OPUS News-Commentary collection in six languages for training and evaluating machine translation models.
A collection of personas for generating diverse synthetic texts and instructions.
550 thousand instruction-response pairs in the style of the IFEval benchmark from Argilla for training models to precisely follow constraints.
A large synthetic dataset of Japanese personas covering demographics, geography, and personality traits.
Cleaned texts of about 9,000 Chinese web novels, split into chapters in JSONL format for LLM training.
A corpus of English mathematical texts of about 200 GB for training and evaluating language models.
500 verified questions and reference SQL queries for SQLite, MySQL, and PostgreSQL simultaneously.
From 10,000 to 100,000 English scientific tasks for comprehensive LLM knowledge evaluation.
A massive multilingual corpus of official UN document translations for training machine translation systems.
Conversational QA dataset: 127k questions with answers from 8k conversations across seven domains.
About 1M text descriptions of US personas generated for LLM training and evaluation.
English long-report and short-summary pairs for training and evaluating summarization models.