The Linear Accessibility Profile predicts steering vector effectiveness and optimal layers with Spearman correlations of 0.86-0.91 using unembedding projections on intermediate states across multiple models and concepts.
Paris” and 20 where it is “London
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Predicting Where Steering Vectors Succeed
The Linear Accessibility Profile predicts steering vector effectiveness and optimal layers with Spearman correlations of 0.86-0.91 using unembedding projections on intermediate states across multiple models and concepts.