Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Rethinking Generalization in Few-Shot Classification
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 |
|---|---|---|---|
| add 15 inner-loop token importance reweighting steps (ablation) | FewTURE ViT-Small · miniImageNet 5-way 5-shot | accuracy: 82.68 -> 84.05 | Helped |
| replace prototype (Euclidean) classifier with reweighted token similarity (ablation) | FewTURE ViT-Small · miniImageNet 5-way 5-shot | accuracy: 82.80 -> 84.05 | Helped |
| replace linear classifier optimized online with reweighted token similarity (ablation) | FewTURE ViT-Small · miniImageNet 5-way 5-shot | accuracy: 82.37 -> 84.05 | Helped |
| use cosine-distance prototypes instead of Euclidean (ablation) | prototype classifier · miniImageNet 5-way 5-shot | accuracy: 82.80 -> 79.90 | Hurt |
| prune patch tokens to 10% using attention (ablation) | FewTURE ViT-Small · miniImageNet 5-way 5-shot | accuracy: 84.05 -> 81.05 | Hurt |
| prune patch tokens to 50% (ablation) | FewTURE ViT-Small · miniImageNet 5-way 5-shot | accuracy: 84.05 -> 83.81 | No clear effect |
| self-supervised MIM pretraining instead of supervised pretraining (no meta fine-tuning) (ablation) | FewTURE backbones · miniImageNet | accuracy: up to 18.19% gain for Swin-tiny | Helped |
| add meta fine-tuning with token reweighting after self-supervised pretraining (ablation) | FewTURE Swin-Tiny · miniImageNet 1-shot | accuracy: +13.01% | Helped |
| add meta fine-tuning with token reweighting after self-supervised pretraining (ablation) | FewTURE ViT-Small · miniImageNet 5-shot | accuracy: +9.94% | Helped |
| add meta fine-tuning to supervised backbones (ablation) | supervised backbones · miniImageNet | accuracy | No clear effect |
| increase reweighting steps from 0 to first steps (ablation) | FewTURE ViT-Small · miniImageNet test | accuracy: +1.15% | Helped |
Closest papers by meaning
- A Closer Look at Few-shot Classification (ICLR 2019 · similarity 0.91)
- Pushing the Limit of Fine-Tuning for Few-Shot Learning: Where Feature Reusing Meets Cross-Scale Attention (AAAI 2024 · similarity 0.89)
- Simple Semantic-Aided Few-Shot Learning (CVPR 2024 · similarity 0.89)
- Unlocking Compositional Generalization in Continual Few-Shot Learning (NeurIPS 2026 · similarity 0.88)
- Provably Improving Generalization of Few-shot models with Synthetic Data (ICML 2025 · similarity 0.88)
- TPR: Topology-Preserving Reservoirs for Generalized Zero-Shot Learning (NeurIPS 2024 · 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.