About this project
Tools for probing and visualising transformer internals — attention patterns, activation patching, and circuit analysis for understanding what neural networks actually learn.
Tech Stack
- Interpretability
- PyTorch
- Transformers
🔬
Mechanistic Interpretability Toolkit
AI Research & ToolingConcept
Tools for probing and visualising transformer internals — attention patterns, activation patching, and circuit analysis for understanding what neural networks actually learn.