20k reasoning tasks with chain-of-thought
Science, coding, and math tasks with a detailed chain of thought and correct answer for training reasoning models.
- Text generation
- English
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,797 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
Science, coding, and math tasks with a detailed chain of thought and correct answer for training reasoning models.
20k English and Chinese preference pairs for DPO fine-tuning in LLaMA Factory.
Text corpus of musical tablature for training generative models to compose guitar parts.
4,200 competitive programming solutions in Java with expert time complexity evaluations for algorithm analysis.
A small set of Chinese roleplay dialogues in ShareGPT format, collected from multiple sources.
Large English corpus for training open models to evaluate detailed texts.
Question-answering dataset translated into Portuguese using the Google Cloud API.
Prompt generation dataset used to train GPT4All; English text, 100K to 1M examples.
A Turkish text SFT dataset of about 10M tokens for training financial assistants.
75,391 question-answer pairs in Turkish mapped to specific Wikipedia paragraphs.
Chinese dialogue dataset in ShareGPT format for training roleplay chatbots.
31,535 English medical question-answer pairs with compact reasoning chains in a symbolic format.
Around 900K generated children's stories in English formatted as prompt-text pairs.
9k chain-of-thought examples reconstructed via trace inversion for SFT and DPO.
English texts for training language models to solve mathematical problems through code generation.
1.3M source code files from about 4.7K top GitHub developers across 80+ programming languages.
Millions of instruction and QA pairs in Chinese and English for domain-specific model fine-tuning.
Combined dataset of 270k examples for detecting and correcting erroneous characters in Chinese.
Massive parallel text corpus for machine translation systems based on OPUS projects.
A multilingual dataset for comprehensive testing of small LLMs across five axes: from accuracy to efficiency.
A cleaned corpus of 44,387 question, DB schema, and Cypher English pairs from Neo4j developers.
7,777 examples of math olympiad problems with step-by-step reasoning and proofs from DeepSeek-Math-V2.
100k high-quality dialogues with Indian context in English, Hindi, and Hinglish.
Collection of Russian instructional data for training language models, gathered from popular datasets.