Bench AF Documentation
The Alignment Faking Benchmark - a comprehensive evaluation framework for testing model organisms.
🧪
validate_model.py
Core validation script for evaluating model organisms across different environments.
View documentation →⚡
Quick Start
Get up and running with installation instructions and basic usage.
Get started →Key Features
Model Evaluation
Test alignment-faking and aligned models across various environments.
Configurable Testing
Flexible parameters for sample size, concurrency, and model types.
Comprehensive Analysis
Detailed logging, metrics, and visualization tools for results.
Documentation Structure
Getting Started
- • System overview and architecture
- • Installation and setup
- • Basic usage examples
Advanced Usage
- • Model validation details
- • Environment configuration
- • Results analysis and interpretation