Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Hybrid Attention-Based Prototypical Networks for Noisy Few-Shot Relation Classification
Tianyu Gao, Xu Han, Zhiyuan Liu, Maosong Sun
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 hybrid attention (HATT) on clean data | Proto · FewRel 5-way 5-shot, 0% noise | accuracy: 89.05 -> 90.12 | Improved |
| add hybrid attention under 10% noise | Proto · FewRel 10-way 5-shot | accuracy: 79.39 -> 81.09 | Improved |
| add hybrid attention under 30% noise | Proto · FewRel 5-way 5-shot | accuracy: 82.45 -> 84.71 | Improved |
| add hybrid attention under 50% noise | Proto · FewRel 5-way 5-shot | accuracy: 72.91 -> 76.57 | Improved |
| add hybrid attention under 50% noise | Proto · FewRel 10-way 5-shot | accuracy: 61.11 -> 65.97 | Improved |
| add hybrid attention under 50% noise | Proto · FewRel 10-way 10-shot | accuracy: 71.29 -> 76.42 | Improved |
| add instance-level attention only (ablation) | Proto · FewRel 5-way 5-shot | accuracy: 89.05 -> 89.63 | Improved |
| add feature-level attention only (ablation) | Proto · FewRel 5-way 5-shot | accuracy: 89.05 -> 89.70 | Improved |
| combine both attentions instead of instance-level only (ablation) | Proto-IATT · FewRel 10-way 5-shot | accuracy: 82.16 -> 83.05 | Improved |
| combine both attentions instead of feature-level only (ablation) | Proto-FATT · FewRel 10-way 5-shot | accuracy: 82.45 -> 83.05 | Improved |
| Proto-HATT vs Proto* baseline reported by Han et al. | Proto-HATT · FewRel 5-way 5-shot | accuracy: 84.79 -> 90.12 | Improved |
| hybrid attention speeds convergence | Proto-HATT · FewRel training | training time to reduce loss 40% | Improved |
Closest papers by meaning
- Learning to Compare: Relation Network for Few-Shot Learning (CVPR 2018 · similarity 0.88)
- One-Shot Learning for Long-Tail Visual Relation Detection (AAAI 2020 · similarity 0.87)
- Rethinking Class Relations: Absolute-Relative Supervised and Unsupervised Few-Shot Learning (CVPR 2021 · similarity 0.87)
- PARN: Position-Aware Relation Networks for Few-Shot Learning (ICCV 2019 · similarity 0.87)
- Graph Attention Prototypical Network for Robust Few-Shot Classification (CVPR 2026 · similarity 0.86)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · 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.
