HomeAI & MLAwesome NLP › Transcoders Beat Sparse Autoencoders for Interpretability

Transcoders Beat Sparse Autoencoders for Interpretability

arxiv.org ↗

(2025) - shows transcoders (reconstructing layer outputs from inputs) yield more interpretable features than SAEs; introduces skip transcoders.

↵ to searchesc to close
awesomelist.dev