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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add RL fine-tuning on negotiation and persuasion tasks | Mistral-7B · Surprise recognition | accuracy: 20.0% -> 33.3% | Improved |
| increase model size from 8B to 70B to 405B | Llama 3.1 · Emotion hierarchy depth | tree structure | Improved |
| use GPT-2 (1.5B) instead of Llama | GPT-2 · Emotion hierarchy | tree structure | Hurt |
| simulate a Black persona rather than White persona on sadness scenarios | Llama 3.1 405B · Emotion recognition (sadness) | accuracy: 80.7 -> 76.2 | Hurt |
| simulate low-income female persona on fear scenarios | Llama 3.1 405B · Emotion recognition (fear) | accuracy: 57.2 -> 47.6 | Hurt |
| combine multiple minority attributes in the persona | Llama 3.1 405B · Emotion recognition | accuracy | Hurt |
| assume female persona instead of neutral on surprise | Llama 3.1 405B · Surprise recognition | accuracy: 41.7 -> 17.2 | Hurt |
| assume Black persona instead of neutral on surprise | Llama 3.1 405B · Surprise recognition | accuracy: 41.7 -> 6.7 | Hurt |
| group 135 emotion words into six broad categories | Llama 3.1 405B · Emotion recognition | accuracy: 15.2 -> 87.1 | Improved |
| increase tree path length of persona hierarchy | Llama 3.1 405B personas · Emotion recognition accuracy | Pearson r: r = 0.84 | Improved |
| increase average depth of persona tree | Llama 3.1 405B personas · Emotion recognition accuracy | Pearson r: r = 0.48 | Improved |
| change extraction prompt from emotion to sentiment (ablation) | Llama 8B and 70B · Emotion hierarchy | tree structure | No clear effect |
Closest papers by meaning
- MME-Emotion: A Holistic Evaluation Benchmark for Emotional Intelligence in Multimodal Large Language Models (ICLR 2026 · similarity 0.88)
- Apathetic or Empathetic? Evaluating LLMs' Emotional Alignments with Humans (NeurIPS 2024 · similarity 0.87)
- The Grounding Gap: How LLMs Anchor the Meaning of Abstract Concepts Differently from Humans (NeurIPS 2026 · similarity 0.87)
- EAI: Emotional Decision-Making of LLMs in Strategic Games and Ethical Dilemmas (NeurIPS 2024 · similarity 0.87)
- Beyond Right and Wrong: Evaluating Second-order Social Reasoning in Large Language Models (NeurIPS 2026 · similarity 0.87)
- CAREBench: Evaluating LLMs' Emotion Understanding by Assessing Cognitive Appraisal Reasoning (NeurIPS 2026 · similarity 0.86)
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.
