Sparse Autoencoders Saes
1 mentions across 0 people
All mentions
Unknown speaker
Recommendedpaper · 2026-06-04
“methods like Sparse Autoencoders (SAEs) may be a promising future direction.”
MLP Activations Are Poor Signals for In-Context Example Selection in LLMs ↗