AI Paper Atlas

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

Do Sparse Autoencoders Identify Reasoning Features in Language Models?

George Ma, Zhongyuan Liang, Irene Y. Chen, Somayeh Sojoudi

ICML 2026 · Large language models › Language, safety and interpretability · approach: Linguistic encoding in LLM representations · 0 citations (OpenAlex)

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
inject a few top-activating tokens into non-reasoning text (ablation)top-100 contrastive SAE reasoning features · 22 configurations (models, layers, datasets)share of features with significant activation increase: 45% to 90% of features activatedImproved
apply LLM-guided counterexample falsification to context-dependent features (ablation)context-dependent SAE features · 248 features across configurationsgenuine reasoning features: 0 of 248 genuineNo clear effect
steer feature 1053 with gamma=2 (ablation)Gemma-3-12B-Instruct layer 22 · GPQA Diamondaccuracy: 37.9% -> 13.6%Hurt
steer feature 1053 with gamma=2 (ablation)Gemma-3-12B-Instruct layer 22 · AIMEaccuracy: 26.7% -> 20.0%Hurt
steer feature 0 with gamma=2 (ablation)Gemma-3-12B-Instruct layer 22 · AIMEaccuracy: 26.7% -> 10.0%Hurt
steer feature 0 with gamma=2 (ablation)Gemma-3-12B-Instruct layer 22 · GPQA Diamondaccuracy: 37.9% -> 20.2%Hurt
steer feature 578 with gamma=2 (ablation)Gemma-3-12B-Instruct layer 22 · AIMEaccuracy: 26.7% -> 26.7%No clear effect
steer feature 578 with gamma=2 (ablation)Gemma-3-12B-Instruct layer 22 · GPQA Diamondaccuracy: 37.9% -> 33.3%Hurt
use Gemma-3-12B layer 22 on s1K instead of other configurationstoken injection analysis · s1Kcontext-dependent fraction: 55% context-dependent vs fewer than 12% in several Gemma configurationsMixed

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