AI Paper Atlas

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

Emotions Where Art Thou: Understanding and Characterizing the Emotional Latent Space of Large Language Models

Benjamin Reichman, Adar Avsian, Larry Heck

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Linguistic encoding in LLM 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
instruction-tuning vs base model (ablation)Llama3.1-8B · emotion alignment to synthetic manifoldcentroid cosine similarity: 0.84 -> 0.93 (English)Improved
instruction-tuning vs base model (ablation)Llama3.1-8B · ML-AURA emotion-selective neuronsneurons per layer separating emotions: +4.8% averageImproved
instruction-tuning vs base model (ablation)Olmov2 · emotion geometry distortionpercentage of high-distortion layers: 47% of layers (English, Instruct)Hurt
instruction-tuning vs base model (ablation)Olmov2 · emotion alignmentStress-2: 0.59 -> 0.32 (English)Improved
project representations into 50D synthetic subspace (ablation)Llama3.1-8B-Base · emotion geometryStress-2: 0.17 -> 0.35Hurt
use English vs non-English datasetsLlama-Base · alignment to synthetic subspacecentroid cosine similarity: 0.84 vs 0.84No clear effect
add learned steering module in emotional subspaceLlama3.1-8B · emotion steering, Englishtop-1 target emotion rate (%): 9 -> 83Improved
add learned steering moduleMinistral · emotion steering, non-Englishtop-1 target emotion rate (%): 11 -> 79Improved
add learned steering moduleOlmov2 · emotion steering, Englishtop-1 target emotion rate (%): 11 -> 91Improved
compare MLP layers to attention layers (ablation)Llama3.1-8B-Base · ML-AURA selectivity% neurons with AUROC > 0.9: 76.5% -> 79%Improved
align synthetic emotion subspace with GSM8K neutral math data (ablation)Llama3.1-8B-Instruct · distortion controlStress-2: 0.284 +- 0.244Hurt

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