Dark Mode
Global LLM Dialogue Dataset
Education & Learning Analytics
Tags and Keywords
Trusted By
"No reviews yet"
Free
About
This dataset is designed for Large Language Model (LLM) training and fine-tuning, particularly for question answering and text generation tasks. It contains over 4 million log and response pairs from three different models across 32 languages, making it a valuable resource for enhancing pre-trained LLMs and improving their performance in various Natural Language Processing (NLP) tasks. The corpus supports instruction tuning and supervised fine-tuning, aiming to improve human language understanding, generate human-like content, and assist in mitigating biases. It is suitable for evaluating LLM capabilities, performing well in classification tasks, and optimising LLM architectures.
Columns
- language: The language in which the prompt was created.
- model: The type of model that generated the response (e.g., GPT-3.5, GPT-4, Uncensored GPT Version).
- time: The timestamp when the model's response was generated.
- text: The user's prompt or query given to the model.
- response: The answer or text generated by the model in response to the prompt.
Distribution
The dataset comprises over 4 million logs/records, typically provided in a CSV file format. It includes log and response pairs generated by three different language models. While specific row counts are not detailed, the substantial number of logs indicates a rich collection for training purposes.
Usage
This dataset is ideal for a range of applications and use cases, including:
- LLM Training: Fine-tuning existing Large Language Models for improved performance.
- Instruction Tuning: Enhancing models to follow specific instructions more effectively.
- Question Answering Systems: Developing and refining systems capable of accurate question answering.
- Text Generation: Creating models that generate human-like and contextually relevant text.
- Text Classification: Training models for various text categorisation tasks.
- NLP Task Improvement: Boosting performance across diverse Natural Language Processing applications.
- LLM Evaluation: Assessing the capabilities and output quality of language models.
- Bias Mitigation: Aiding in the reduction of biases within LLM outputs.
- LLM Architecture Optimisation: Supporting the development of more effective language processing architectures.
Coverage
The dataset is global in its scope, featuring logs written in 32 different languages, including but not limited to English, Chinese, Arabic, French, German, Japanese, Korean, Portuguese, Spanish, and Turkish. The data spans a time range from April 2023 to January 2024, offering recent language model interactions.
License
CC-BY-NC
Who Can Use It
This dataset is suitable for:
- AI/ML Researchers: For academic studies on LLM behaviour, fine-tuning, and performance.
- Data Scientists: To build and improve NLP models and applications.
- LLM Developers: For instruction tuning, supervised fine-tuning, and optimising custom language models.
- NLP Engineers: To enhance text generation capabilities, refine question answering, and develop classification systems.
- Organisations focused on AI: To develop and deploy robust, high-performing language processing solutions.
Dataset Name Suggestions
- LLM Fine-Tuning Question Answering Dataset
- Multilingual AI Text Generation Log
- Language Model Instruction Tuning Corpus
- Global LLM Dialogue Data
- NLP Model Response Archive
Attributes
Original Data Source: LLM Text Generation Dataset
QUALITY