AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Domain Restriction via SAE Multi-Layer Transitions

Elias Shaheen, Avi Mendelson

ICML 2026 · Large language models › Language, safety and interpretability · approach: Mechanistic analysis of attention and representations

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use SAE features instead of raw activations (ablation)HTM scorer · AGNews near-OOD (mean of 4 splits), Gemma2-2BAUROC: 0.753 -> 0.912Improved
use first-order Markov scorer instead of HTM (ablation)SAE-SDR sequence scorer · AGNews near-OOD, Gemma2-2BAUROC: 0.912 -> 0.919No clear effect
use LSTM predictor instead of Markov (ablation)SAE-SDR sequence scorer · AGNews near-OOD, Gemma2-2BAUROC: 0.919 -> 0.910No clear effect
add 1-hop layer transitions to static registry scoring (ablation)trajectory registry scoring · AGNews pairs, mean over 12AUROC: 0.8445 -> 0.9159Improved
extend registry from 1-hop to 2-hop (ablation)trajectory registry scoring · AGNews pairs, mean over 12AUROC: 0.9159 -> 0.9128No clear effect
increase Top-k from 10 to 128 (ablation)Markov scorer · AGNews, Gemma2-2BAUROC: 0.9280 -> 0.9076Hurt
increase Top-k from 10 to 128 (ablation)HTM scorer · AGNews, Gemma2-2BAUROC: 0.9238 -> 0.5000Hurt
use Gemma2-9B instead of 2BSAE+Markov · 20NG vs IMDB far-OODAUROC: 0.9171 -> 0.9540Improved
use Gemma2-9B instead of 2BSAE+Markov · CLINC150 near-OODAUROC: 0.7387 -> 0.8364Improved
use Gemma2-9B instead of 2BSAE+Markov · SNIPS near-OODAUROC: 0.9699 -> 0.9001Hurt
use SAE-transition method instead of likelihood-ratio baselineGemma2-2B · 20NG vs SST-2 far-OODAUROC: 0.9988 -> 0.9782Hurt
use SAE-transition method instead of likelihood-ratio baselineGemma2-2B · AGNews near-OOD (class c3)AUROC: 0.76 -> 0.98Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy