flâneur — a map of the web's best reading

Negative Results for SAEs On Downstream Tasks and Deprioritising SAE Research (GDM Mech Interp Team Progress Update #2) — LessWrong

lesswrong.com · 13,828 words · saved by 1 readers

Lewis Smith*, Sen Rajamanoharan*, Arthur Conmy, Callum McDougall, Janos Kramar, Tom Lieberum, Rohin Shah, Neel Nanda • * = equal contribution …

x Negative Results for SAEs On Downstream Tasks and Deprioritising SAE Research (GDM Mech Interp Team Progress Update #2) — LessWrong Interpretability (ML & AI) Sparse Autoencoders (SAEs) AI Curated 2025 Top Fifty: 39 % 117 Negative Results for SAEs On Downstream Tasks and Deprioritising SAE Research (GDM Mech Interp Team Progress Update #2) by lewis smith , Senthooran Rajamanoharan , Arthur Conmy , CallumMcDougall , Tom Lieberum , János Kramár , Rohin Shah , Neel Nanda 26th Mar 2025 AI Alignment Forum Linkpost for deepmindsafetyresearch.medium.com 35 min read 15 117 Ω 58 Lewis Smith*, Sen Raj

Explore this link on the map →

related reading