🦙🛁 Cleaned Alpaca Dataset

This repository hosts a cleaned and curated version of the dataset used to train the Alpaca LLM. On April 8, 2023, ~50,000 uncurated instructions were replaced with GPT-4-LLM data. Curation is ongoing.

7B and 13B LoRA models (trained in April 2023) are available on Hugging Face:

• 7B Model
• 13B Model

High-quality data improves model performance, often more effectively than increasing model size.

🧹 Data Cleaning and Curation

The original GPT-3-generated dataset had issues like noise, bias, and poor loss curves. The cleaned version addresses these, improving performance and reducing hallucinations.

Key Issues Fixed:

• Noisy and inconsistent data.
• US-centric bias.
• Over-reliance on GPT-3 limitations.

🚀 Applications

Used in:

• Multilingual chatbots.
• Educational and healthcare tools.
• Creative writing and research assistance.

🔮 Future Plans

• Expand cultural diversity.
• Incorporate real-time updates.
• Integrate user feedback.

🤝 Contribute

Help by:

• Submitting data.
• Reporting bugs.
• Improving documentation.

🌟 Success Stories

• Startups improved chatbot accuracy by 30%.
• Universities reduced faculty workload by 20%.
• Non-profits built multilingual support tools.