joelburget / SAELens

Training Sparse Autoencoders on Language Models

SAE Lens

SAELens exists to help researchers:

• Train sparse autoencoders.
• Analyse sparse autoencoders / research mechanistic interpretability.
• Generate insights which make it easier to create safe and aligned AI systems.

Please refer to the documentation for information on how to:

• Download and Analyse pre-trained sparse autoencoders.
• Train your own sparse autoencoders.
• Generate feature dashboards with the SAE-Vis Library.

SAE Lens is the result of many contributors working collectively to improve humanities understanding of neural networks, many of whom are motivated by a desire to safeguard humanity from risks posed by artificial intelligence.

This library is maintained by Joseph Bloom and David Channin.

Tutorials

• Loading and Analysing Pre-Trained Sparse Autoencoders
• Understanding SAE Features with the Logit Lens
• Training a Sparse Autoencoder

Join the Slack!

Feel free to join the Open Source Mechanistic Interpretability Slack for support!

Citations and References

Research:

• Towards Monosemanticy
• Sparse Autoencoders Find Highly Interpretable Features in Language Model

Reference Implementations:

• Neel Nanda
• AI-Safety-Foundation.
• Arthur Conmy.
• Callum McDougall