Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Attribute Attention for Semantic Disambiguation in Zero-Shot Learning
Yang Liu, Jishun Guo, Deng Cai, Xiaofei He
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 attribute attention (ablation) | LFGAA+Hybrid · AwA2 (proposed split) | accuracy: 62.4 -> 68.1 | Improved |
| add attribute attention (ablation) | LFGAA+Hybrid · CUB (proposed split) | accuracy: 62.1 -> 67.6 | Improved |
| add attribute attention (ablation) | LFGAA+Hybrid · SUN (proposed split) | accuracy: 61.2 -> 61.5 | No clear effect |
| add attribute attention (ablation) | LFGAA+SA · AwA2 (proposed split) | accuracy: 70.9 -> 75.5 | Improved |
| add attribute attention (ablation) | LFGAA+SA · CUB (proposed split) | accuracy: 67.4 -> 78.9 | Improved |
| add attribute attention (ablation) | LFGAA+SA · SUN (proposed split) | accuracy: 62.8 -> 66.2 | Improved |
| use softmax attention instead of sigmoid attention (ablation) | LFGAAG+SA · CUB (proposed split) | accuracy: 73.5 -> 81.1 | Improved |
| use softmax attention instead of no attention (ablation) | LFGAAG+SA · CUB (proposed split) | accuracy: 62.1 -> 81.1 | Improved |
| use softmax attention instead of no attention (ablation) | LFGAAG+SA · CUB (proposed split) | prediction variance: 2.48 -> 0.86 | Improved |
| use self-adapted prototypes (SA) instead of hybrid prototypes (ablation) | LFGAA · CUB (proposed split) | accuracy: 67.6 -> 78.9 | Improved |
| use transductive self-adaptation (SA) instead of hybrid (ablation) | LFGAA · SUN (proposed split) | accuracy: 61.5 -> 66.2 | Improved |
| use LFGAA+SA over QFSL | LFGAA+SA · CUB (proposed split, transductive) | accuracy: 72.1 -> 81.1 | Improved |
Closest papers by meaning
- Mutually Causal Semantic Distillation Network for Zero-Shot Learning (IJCV 2026 · similarity 0.91)
- Learning Discriminative Latent Attributes for Zero-Shot Classification (ICCV 2017 · similarity 0.91)
- Semantics Disentangling for Generalized Zero-Shot Learning (ICCV 2021 · similarity 0.91)
- DUET: Cross-Modal Semantic Grounding for Contrastive Zero-Shot Learning (AAAI 2023 · similarity 0.91)
- SAGE: Structured Attribute-Guided Enhancement for GZSL (AAAI 2026 · similarity 0.90)
- ZeroDiff: Solidified Visual-semantic Correlation in Zero-Shot Learning (ICLR 2025 · similarity 0.89)
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.
