Research map / Vision-language and multimodal / Human and embodied modeling
Affection: Learning Affective Explanations for Real-World Visual Data
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Fine-tune BERT rather than train an LSTM text classifier from scratch | Affection text-to-emotion classification · Affection test | 9-way emotion accuracy: 69.8% -> 72.5% | Helped |
| Use an Affection-trained neural listener instead of pretrained CLIP | Image retrieval from affective explanations · Affection image retrieval (1 distractor) | retrieval accuracy: 91.1% vs. 97.9% | Hurt |
| Use an Affection-trained neural listener instead of pretrained CLIP | Image retrieval from affective explanations · Affection image retrieval (10 distractors) | retrieval accuracy: 63.4% vs. 89.7% | Hurt |
| Add emotion-label conditioning (ablation) | SAT affective speaker · Affection test | Emo-Alignment: 48.1 -> 55.2 | Helped |
| Add emotion-label conditioning (ablation) | SAT affective speaker · Affection test | BLEU-1: 64.4 -> 63.1 | Hurt |
| Add CLIP-based pragmatic reranking (ablation) | Default SAT affective speaker · Affection test | CLIPScore: 66.7 -> 69.2 | Helped |
| Add CLIP-based pragmatic reranking (ablation) | Default SAT affective speaker · Affection test | Unique-Productions: 78.7 -> 82.9 | Helped |
| Add CLIP-based pragmatic reranking (ablation) | Emo-Grounded SAT affective speaker · Affection test | Emo-Alignment: 55.2 -> 55.9 | No clear effect |
| Add CLIP-based pragmatic reranking (ablation) | Emo-Grounded SAT affective speaker · Affection test | Emotional Turing test aggregate success rate: 59.6% -> 65.2% | Helped |
Closest papers by meaning
- Customizing Visual Emotion Evaluation for MLLMs: An Open-vocabulary, Multifaceted, and Scalable Approach (ICLR 2026 · similarity 0.89)
- EmoEdit: Evoking Emotions through Image Manipulation (CVPR 2025 · similarity 0.88)
- AffectGPT: A New Dataset, Model, and Benchmark for Emotion Understanding with Multimodal Large Language Models (ICML 2025 · similarity 0.88)
- Bridge Then Begin Anew: Generating Target-Relevant Intermediate Model for Source-Free Visual Emotion Adaptation (AAAI 2025 · similarity 0.88)
- EmoNet-Face: An Expert-Annotated Benchmark for Synthetic Emotion Recognition (NeurIPS 2025 · similarity 0.88)
- VidEmo: Affective-Tree Reasoning for Emotion-Centric Video Foundation Models (NeurIPS 2025 · 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.