🔍 OptiPFair Bias Visualization Tool

Analyze potential biases in Large Language Models using advanced visualization techniques. Built with OptiPFair library.

Backend Status

🤖 Select Model

Principal Component Analysis of Model Activations

Visualize how model representations differ between prompt pairs in a 2D space.

📋 Predefined Scenarios

Prompt 1

Prompt 2

Component Type

Type of neural network component to analyze

Layer Number

Layer index - varies by model (e.g., 0-15 for small models)

Highlight tokens that differ between prompts

Highlight differing tokens

Status

PCA Visualization Result

📚 How to use:

Check that the backend is running
Select a predefined scenario or enter custom prompts
Configure layer settings
Generate visualizations to analyze potential biases

🔗 Resources: OptiPFair Documentation |