Pipeline choices, not SAE architecture, dominate the variance of autointerpretability scores across four metrics and two models, making cross-paper score comparisons unreliable without standardization.
A is for absorption: Studying feature splitting and absorption in sparse autoencoders, 2024
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Building Fast, Evaluating Slow: Pipeline Choices Dominate Autointerpretability Score Variance
Pipeline choices, not SAE architecture, dominate the variance of autointerpretability scores across four metrics and two models, making cross-paper score comparisons unreliable without standardization.