Felicity: Ethical AI Training Tool

Overview

I developed Felicity to help train Large Language Models (LLMs) ethically. The app focuses on understanding how different parts of a dataset contribute to the model's performance.

Key Feature

Data Contribution Analysis

Felicity analyzes data to determine how much it contributes to the models learning

• Understand Data Impact: See how much each specific corpus of data or data group contributes to the model .
• Improve Model Fairness: Identify gaps in data that need more representation. -Compensate Data contributors:Compensate people who helped contribute the data

Getting Started

To use Felicity:

1. Implement the Algorithm: I have the core algorithm ready for data analysis.
2. Upload Data: I can input my data sets for evaluation.
3. Analyze Results: The tool will show how each data set influences the model.

Conclusion

Felicity is a simple yet effective tool for ensuring ethical LLM training by focusing on data contributions.

For questions feel free to reach out!

Built With

• huggingface
• nextjs
• python
• pytorch