Research map / Video / Video-language understanding
Benchmarking Dynamic Affective Reasoning: A Viewer-Centric Video Emotion Dataset
Zhiyan Zhang, Peipei Song, Jinpeng Hu, Jingyang Jia, Xun Yang, Xiaojun Chang
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 SFT on DAR to Qwen2.5-VL-3B (ablation) | Qwen2.5-VL-3B · DAR test set | Emo-Acc: 15.0 -> 25.8 | Improved |
| add GRPO stage with multi-component rewards after SFT (ablation) | DAR-SFT · DAR test set | mIoU: 47.6 -> 52.3 | Improved |
| add GRPO stage after SFT (ablation) | DAR-SFT · DAR test set | Emo-Acc: 25.8 -> 28.6 | Improved |
| add GRPO stage after SFT (ablation) | DAR-SFT · DAR test set | GPT reasoning score avg: 3.0 -> 3.3 | Improved |
| add segment-count reward R_count (ablation) | DAR-R1 · DAR test set | SC-Acc: 38.3 -> 41.5 | Improved |
| add segmentation reward R_seg (ablation) | DAR-R1 · DAR test set | mIoU: 48.0 -> 52.3 | Improved |
| add emotion accuracy reward R_emo (ablation) | DAR-R1 · DAR test set | Emo-Acc: 24.2 -> 28.6 | Improved |
| add reasoning quality reward R_reason (ablation) | DAR-R1 · GRPO training | training stability | Improved |
| fine-tune emotion-focused MLLMs on DAR instead of DAR-R1 training | AffectGPT · DAR test set | Emo-Acc: AffectGPT-FT 6.6 vs DAR-R1 28.6 | Improved |
| DAR-R1 evaluated out-of-domain on TSL | DAR-R1 · TSL | mAP@0.1: 7.9 (AffectGPT) -> 13.2 | Improved |
Closest papers by meaning
- VidEmo: Affective-Tree Reasoning for Emotion-Centric Video Foundation Models (NeurIPS 2025 · similarity 0.90)
- $E^3$: Exploring Embodied Emotion Through A Large-Scale Egocentric Video Dataset (NeurIPS 2024 · similarity 0.88)
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (ICML 2025 · similarity 0.88)
- Customizing Visual Emotion Evaluation for MLLMs: An Open-vocabulary, Multifaceted, and Scalable Approach (ICLR 2026 · similarity 0.86)
- Facial-R1: Aligning Reasoning and Recognition for Facial Emotion Analysis (AAAI 2026 · similarity 0.86)
- Uncertain Multimodal Intention and Emotion Understanding in the Wild (CVPR 2025 · 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.
