In an additive residual-stream model with a linear readout, ablating a subset of carriers collapses a matched input pair exactly when the removed mass is symmetric and no contrast remains outside the subset; patching and ablation move readouts by different quantities.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
A Theory of Conditional Collapse under Low-Rank Weight-Space Ablations: I. The Single-Block Theory and Synthetic Validation
In an additive residual-stream model with a linear readout, ablating a subset of carriers collapses a matched input pair exactly when the removed mass is symmetric and no contrast remains outside the subset; patching and ablation move readouts by different quantities.