Where this topic leads
Topics that build on Sparse Autoencoders for Interpretability: TopK, JumpReLU, Matryoshka, and Scaling
Once you have Sparse Autoencoders for Interpretability: TopK, JumpReLU, Matryoshka, and Scaling, these are the topics that cite it as a prerequisite. Pick by tier and the area you want to push into next.
Editor's suggested next (3)
Core flagship topics (1)
- Mechanistic Interpretability: Features, Circuits, and Causal Faithfulnesslayer 4 · ai-safety