AI Paper Atlas

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

From Associations to Activations: Comparing Behavioral and Hidden-State Semantic Geometry in LLMs

Louis Schiekiera, Max Zimmer, Christophe Roux, Sebastian Pokutta, Fritz Günther

ICML 2026 · Large language models › Language, safety and interpretability · approach: Reasoning mechanisms and failure modes · 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
use forced choice instead of free association behavior (ablation)behavioral similarity · RSA with hidden statesPearson r: .18 -> .43Improved
add FC similarity to baseline predictors (ablation)ridge regression · held-out-words hidden-state similaritytest R2: +.022Improved
add FA similarity to baseline predictors (ablation)ridge regression · held-out-words hidden-state similaritytest R2: +.002No clear effect
extract hidden states with task/meaning prompt instead of averaging over C4 sentences (ablation)RSA with FC · RSAmean Pearson r: .35 -> .46Improved
use Averaged extraction instead of meaning-focused prompts (ablation)RSA with FC · layer of peak alignmentpeak depth: about 30% -> about 72% depthMixed
use FC instead of FA by semantic category (ablation)RSA · 20 of 21 WordNet supersensesPearson r: delta r about +.30 abstract, +.13 concreteImproved
increase cue-word frequency (ablation)FC alignment · RSAPearson r: .37 -> .47Improved
use cross-model consensus as predictor instead of behaviorhidden-state similarity prediction · RSAPearson r: .43 -> .74Improved
use FC behavior instead of FastTextbehavioral reference · RSAPearson r: .20 -> .43Improved
increase neighborhood size k for FC (ablation)nearest-neighbor overlap · NN@kmean overlap: .20 -> .29Improved

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