Research map / Vision-language and multimodal / Multimodal LLMs
Facial Affective Behavior Analysis with Instruction Tuning
Yifan Li, Anh Dao, Wentao Bao, Zhen Tan, Tianlong Chen, Huan Liu, Yu Kong
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 facial landmark prior token alongside visual tokens (ablation) | EmoLA · FABA-Bench emotion | recognition accuracy S_re: 62.5 -> 64.5 | Improved |
| add facial landmark prior token alongside visual tokens (ablation) | EmoLA · FABA-Bench AU | recognition F1 S_re: 55.3 -> 56.3 | Improved |
| add visual tokens to the prior-only input (ablation) | EmoLA · FABA-Bench emotion | recognition accuracy S_re: 41.2 -> 62.5 | Improved |
| tune LoRA with visual projector in addition to prior projector (ablation) | EmoLA · FABA-Bench emotion | recognition accuracy S_re: 44.9 -> 64.5 | Improved |
| also tune prior projector on top of LoRA and visual projector (ablation) | EmoLA · FABA-Bench AU | recognition F1 S_re: 55.6 -> 56.3 | Improved |
| multi-task training on emotion and AU instead of single task (ablation) | EmoLA · FABA-Bench AU | REGE: 91.5 -> 88.6 | Hurt |
| multi-task training on emotion and AU instead of single task (ablation) | EmoLA · FABA-Bench emotion | REGE: 96.2 -> 96.3 | No clear effect |
| place prior token after visual tokens rather than before (ablation) | EmoLA · FABA-Bench AU | REGE: 89.7 -> 91.5 | Improved |
| use EmoLA (LoRA, facial prior) instead of fully fine-tuned LLaVA-1.5 | FABA-Bench emotion · FABA-Bench emotion | REGE: 93.9 -> 96.2 | Improved |
| use EmoLA instead of ReCoT on BP4D | BP4D · BP4D | average F1 (%): 64.8 -> 64.2 | Hurt |
Closest papers by meaning
- Facial-R1: Aligning Reasoning and Recognition for Facial Emotion Analysis (AAAI 2026 · similarity 0.89)
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (ICML 2025 · similarity 0.86)
- Customizing Visual Emotion Evaluation for MLLMs: An Open-vocabulary, Multifaceted, and Scalable Approach (ICLR 2026 · similarity 0.86)
- MME-Emotion: A Holistic Evaluation Benchmark for Emotional Intelligence in Multimodal Large Language Models (ICLR 2026 · similarity 0.85)
- Emotion-LLaMA: Multimodal Emotion Recognition and Reasoning with Instruction Tuning (NeurIPS 2024 · similarity 0.85)
- EmoNet-Face: An Expert-Annotated Benchmark for Synthetic Emotion Recognition (NeurIPS 2025 · similarity 0.85)
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.
