Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
A Shared Multi-Attention Framework for Multi-Label Zero-Shot Learning
Check 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 |
|---|---|---|---|
| use shared multi-attention (M = 10) instead of Fast0Tag | LESA · NUS-WIDE GZS | F1@3: improves the F1 score of the state of the art by 2.9% | Helped |
| use shared multi-attention (M = 10) instead of Fast0Tag | LESA · Open Images GZS | F1@10: 1.4% | Helped |
| add the region relevance loss to ranking loss (ablation) | LESA · NUS-WIDE ZS | mAP: 13.5 -> 16.9 | Helped |
| add the region relevance loss to ranking loss (ablation) | LESA · NUS-WIDE ZS | F1@3: 28.3 -> 31.0 | Helped |
| add the diversity loss to ranking and relevance losses (ablation) | LESA · NUS-WIDE ZS | mAP: 16.9 -> 18.0 | Helped |
| add the diversity loss to ranking and relevance losses (ablation) | LESA · NUS-WIDE ZS | F1@3: 31.0 -> 31.3 | Helped |
| add the attention distribution loss to ranking, relevance, and diversity losses (ablation) | LESA · NUS-WIDE ZS | mAP: 18.0 -> 19.4 | Helped |
| use one attention module instead of ten (ablation) | LESA · NUS-WIDE ZS | F1@3: 29.8 -> 31.6 | Hurt |
| assign one attention module per semantic-vector cluster (ablation) | LESA · NUS-WIDE ZS | F1@3: 24.6 -> 31.6 | Hurt |
| increase the number of attention modules beyond 10 (ablation) | LESA · NUS-WIDE | F1 | Hurt |
| share attention modules across labels rather than learn one per label | LESA · Open Images test localization | Harmonic mean (seen+unseen): 3.4 -> 4.2 | Helped |
| use one attention for all labels instead of shared multiple attentions (ablation) | LESA · Open Images test localization | Harmonic mean (seen+unseen): 3.8 -> 4.2 | Hurt |
Closest papers by meaning
- Saliency-based Sequential Image Attention with Multiset Prediction (NeurIPS 2017 · similarity 0.88)
- Unlocking the Power of Co-Occurrence in CLIP: A DualPrompt-Driven Method for Training-Free Zero-Shot Multi-Label Classification (ICLR 2026 · similarity 0.88)
- Epsilon: Exploring Comprehensive Visual-Semantic Projection for Multi-Label Zero-Shot Learning (AAAI 2025 · similarity 0.87)
- Correlative and Discriminative Label Grouping for Multi-Label Visual Prompt Tuning (CVPR 2025 · similarity 0.86)
- Attention-Induced Embedding Imputation for Incomplete Multi-View Partial Multi-Label Classification (AAAI 2024 · similarity 0.86)
- Language-Driven Cross-Modal Classifier for Zero-Shot Multi-Label Image Recognition (ICML 2024 · 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.