Chat Instruct EN
This dataset brings together thousands of instructions and responses generated by LLM models, in the form of dialogues formatted for training conversational agents.
JSONL format, several tens of thousands of exchanges (size ≈ not specified)
CC BY 4.0
Description
Chat Instruct EN is a textual corpus in English, consisting of examples in instruction—response format. It aggregates several public datasets like Infinity-Instruct and LMSYS-Chat, cleaned and homogenized. Each line in the JSONL file contains a user-model interaction, useful for training conversational agents or for fine-tuning models on instruction-guided tasks.
What is this dataset for?
- Refine an LLM to better follow specific instructions (tuning instruction)
- Create or train contextual chatbots
- Test the response and dialogue capabilities of AI models
Can it be enriched or improved?
Yes. You can enrich this dataset by adding metadata (type of task, difficulty, tone), translations into other languages, or multi-turn variants. It can also be segmented by type of prompt (definition, explanation, practical task) for more targeted training.
🔎 In summary
🧠 Recommended for
- Chatbot developers
- LLM researchers
- Prompt engineering testers
🔧 Compatible tools
- Hugging Face Transformers
- OpenChatKit
- LangChain
- FastChat
💡 Tip
Extract the prompts alone to generate new examples or train a model to complete the dialog.
Frequently Asked Questions
Does this dataset contain multi-turn conversations?
No, each entry corresponds to a single instruction followed by an answer, in the simple “user/assistant” format.
Can we filter the dataset according to the type of task?
Not directly, but pre-processing makes it possible to identify instructions such as definition, generation, classification, etc.
Is the dataset multilingual?
No, all content has been automatically filtered to include only examples in English.
