NeurIPS 2025 Explorer
Concepts
Authors
Glossary
johnsanterre.github.io
latent directions
3 papers
A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencoders
A is for Absorption: Studying Feature Splitting and Absorption in Sparse Autoencoders
Overcoming Sparsity Artifacts in Crosscoders to Interpret Chat-Tuning