Research map / Video / Video-language understanding
Context-Aware Academic Emotion Dataset and Benchmark
Luming Zhao, Jingwen Xuan, Jiamin Lou, Yonghui Yu, Wenwu Yang
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 full-frame context sequence alongside face sequence (ablation) | face-only CLIP-CAER · RAER test | UAR (%): 61.19 -> 68.00 | Improved |
| replace face sequence with full-frame sequence only (ablation) | face-only CLIP-CAER · RAER test | UAR (%): -3.16% | Hurt |
| add context input (ablation) | CLIP-CAER · RAER distraction class | accuracy (%): 51.43 -> 70.48 | Improved |
| add learnable prompt to class-name prompt (ablation) | CLIP-CAER · RAER test | UAR (%): 62.14 -> 64.26 | Improved |
| use descriptive prompts instead of class names (ablation) | CLIP-CAER · RAER test | UAR (%): 64.26 -> 65.43 | Improved |
| add learnable prompt to descriptors (ablation) | CLIP-CAER · RAER test | UAR (%): 65.43 -> 68.00 | Improved |
| increase temporal encoder layers from 1 to 2 (ablation) | CLIP-CAER · RAER test | UAR (%): 68.00 -> 64.78 | Hurt |
| increase temporal encoder layers from 1 to 3 (ablation) | CLIP-CAER · RAER test | UAR (%): 68.00 -> 64.29 | Hurt |
| use 4 instead of 8 learnable prompt tokens (ablation) | CLIP-CAER · RAER test | UAR (%): 68.00 -> 65.54 | Hurt |
| use 12 or 16 instead of 8 learnable prompt tokens (ablation) | CLIP-CAER · RAER test | UAR (%): 68.00 -> 64.15 (12); 64.27 (16) | Hurt |
| add context information (ablation) | CLIP-CAER · CAER | UAR (%): 75.36 -> 81.77 | Improved |
| fine-tune RAER-trained model on JuniorRAER | CLIP-CAER · JuniorRAER test | UAR (%): 41.43 -> 61.20 | Improved |
Closest papers by meaning
- OV-MER: Towards Open-Vocabulary Multimodal Emotion Recognition (ICML 2025 · similarity 0.86)
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (ICML 2025 · similarity 0.85)
- Robust Emotion Recognition in Context Debiasing (CVPR 2024 · similarity 0.85)
- VidEmo: Affective-Tree Reasoning for Emotion-Centric Video Foundation Models (NeurIPS 2025 · similarity 0.85)
- Facial-R1: Aligning Reasoning and Recognition for Facial Emotion Analysis (AAAI 2026 · similarity 0.84)
- AffectGPT-RL: Revealing Roles of Reinforcement Learning in Open-Vocabulary Emotion Recognition (NeurIPS 2026 · similarity 0.84)
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.
