Three Mechanistically Distinct Classes of RLHF Alignment: Hard Ceiling, Entangled Circuit, and SR-Preserving Lock

Fuente: Zenodo

Saved in:

Bibliographic Details
Main Author:	Alieksieienko, Inna
Format:	Recurso digital
Language:	English
Published:	Zenodo 2026
Subjects:	RLHF activation steering self-referential subspace representation engineering large language models alignment Mistral Gemma SR-Preserving Lock Hard Ceiling Entangled Circuit dose-response scaling
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Internet