Research map / Large language models / Alignment and preferences
EmoPrefer: Can Large Language Models Understand Human Emotion Preferences?
Zheng Lian, Licai Sun, Lan Chen, Haoyu Chen, Zebang Cheng, Fan Zhang, Ziyu Jia, Ziyang Ma, Fei Ma, Xiaojiang Peng, Jianhua Tao
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 |
|---|---|---|---|
| use external LLM (Qwen2.5-7B) for the preference step (S3/S4) instead of the MLLM itself (S1/S2) (ablation) | most open-source MLLMs · EmoPrefer-Data | WAF (2-class): LLaMA-VID 33.03 (S1) -> 57.10 (S4); VideoChat 38.81 -> 51.79 | Improved |
| use S3/S4 external-LLM pipeline for strong MLLMs (ablation) | Qwen2.5-VL, Qwen2.5-Omni · EmoPrefer-Data | WAF (2-class): Qwen2.5-VL 64.43 (S1) -> 61.79 (S3); Qwen2.5-Omni 67.21 (S2) -> 61.54 (S3) | Hurt |
| add reasoning sub-step (S4 over S3) (ablation) | weaker open-source MLLMs · EmoPrefer-Data | WAF (2-class): VideoChat 45.94 -> 51.79; Qwen2-Audio 58.10 -> 63.17 | Improved |
| use larger external LLM (Qwen3-14B instead of Qwen2.5-7B) (ablation) | S3/S4 judge pipeline · EmoPrefer-Data | WAF (2-class) | No clear effect |
| aggregate normal and swapped order predictions by majority vote (ablation) | single MLLM judges · EmoPrefer-Data | WAF (2-class) | Mixed |
| model-based crowdsourcing: aggregate top-k MLLM judges (ablation) | MLLM judges · EmoPrefer-Data | WAF and swap consistency | Improved |
| add normal-swapped combination within crowdsourcing (ablation) | model-based crowdsourcing · EmoPrefer-Data | WAF | No clear effect |
| restrict crowdsourcing to closed-source models only (ablation) | model-based crowdsourcing · EmoPrefer-Data | WAF | Hurt |
Closest papers by meaning
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (ICML 2025 · similarity 0.90)
- OV-MER: Towards Open-Vocabulary Multimodal Emotion Recognition (ICML 2025 · similarity 0.90)
- Customizing Visual Emotion Evaluation for MLLMs: An Open-vocabulary, Multifaceted, and Scalable Approach (ICLR 2026 · similarity 0.89)
- MME-Emotion: A Holistic Evaluation Benchmark for Emotional Intelligence in Multimodal Large Language Models (ICLR 2026 · similarity 0.89)
- Emotion-Coherent Reasoning for Multimodal LLMs via Emotional Rationale Verifier (AAAI 2026 · similarity 0.89)
- Emotion-LLaMA: Multimodal Emotion Recognition and Reasoning with Instruction Tuning (NeurIPS 2024 · similarity 0.88)
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.
