residual-paving

Tag

Cards List
#residual-paving

Residual Paving: Diagnosing the Routing Bottleneck in Selective Refusal Editing

arXiv cs.LG · 2026-05-21 Cached

Introduces Residual Paving, a routed residual editing method for frozen instruction-tuned transformers that separates route selectivity from residual-edit capacity, achieving significant reduction in edit refusal on Gemma-3-4B-IT while preserving benign and harmful behavior.

0 favorites 0 likes
← Back to home

Submit Feedback