AI Paper Atlas

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

Emergence of Hierarchical Emotion Organization in Large Language Models

Maya Okawa, Bo Zhao, Eric Bigelow, Rose Yu, Tomer Ullman, Ekdeep Singh Lubana, Hidenori Tanaka

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
add RL fine-tuning on negotiation and persuasion tasksMistral-7B · Surprise recognitionaccuracy: 20.0% -> 33.3%Improved
increase model size from 8B to 70B to 405BLlama 3.1 · Emotion hierarchy depthtree structureImproved
use GPT-2 (1.5B) instead of LlamaGPT-2 · Emotion hierarchytree structureHurt
simulate a Black persona rather than White persona on sadness scenariosLlama 3.1 405B · Emotion recognition (sadness)accuracy: 80.7 -> 76.2Hurt
simulate low-income female persona on fear scenariosLlama 3.1 405B · Emotion recognition (fear)accuracy: 57.2 -> 47.6Hurt
combine multiple minority attributes in the personaLlama 3.1 405B · Emotion recognitionaccuracyHurt
assume female persona instead of neutral on surpriseLlama 3.1 405B · Surprise recognitionaccuracy: 41.7 -> 17.2Hurt
assume Black persona instead of neutral on surpriseLlama 3.1 405B · Surprise recognitionaccuracy: 41.7 -> 6.7Hurt
group 135 emotion words into six broad categoriesLlama 3.1 405B · Emotion recognitionaccuracy: 15.2 -> 87.1Improved
increase tree path length of persona hierarchyLlama 3.1 405B personas · Emotion recognition accuracyPearson r: r = 0.84Improved
increase average depth of persona treeLlama 3.1 405B personas · Emotion recognition accuracyPearson r: r = 0.48Improved
change extraction prompt from emotion to sentiment (ablation)Llama 8B and 70B · Emotion hierarchytree structureNo clear effect

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