Local linearity of LLM layers enables LQR-based closed-loop activation steering with theoretical tracking guarantees.
Pawan , year=
2 Pith papers cite this work, alongside 18 external citations. Polarity classification is still indexing.
2
Pith papers citing it
18
external citations · external index
citation-role summary
method 1
citation-polarity summary
years
2026 2roles
method 1polarities
use method 1representative citing papers
The paper introduces RoMA and gRoMA as statistical tools that compute auditable upper bounds on the failure probability of any black-box AI system once regulators fix an acceptable risk threshold and input domain.
citing papers explorer
-
Local Linearity of LLMs Enables Activation Steering via Model-Based Linear Optimal Control
Local linearity of LLM layers enables LQR-based closed-loop activation steering with theoretical tracking guarantees.
-
Bounding the Black Box: A Statistical Certification Framework for AI Risk Regulation
The paper introduces RoMA and gRoMA as statistical tools that compute auditable upper bounds on the failure probability of any black-box AI system once regulators fix an acceptable risk threshold and input domain.