Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
BatchFormer: Learning To Explore Sample Relationships for Robust Representation 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 |
|---|---|---|---|
| Add BatchFormer to Balanced Softmax | CIFAR-100-LT · CIFAR-100-LT, imbalance ratio 100 | All accuracy: 50.7 -> 51.7 | Helped |
| Add BatchFormer to Balanced Softmax | CIFAR-100-LT · CIFAR-100-LT, imbalance ratio 200 | All accuracy: 46.4 -> 47.5 | Helped |
| Add BatchFormer | Balanced Softmax, ResNet-10 · ImageNet-LT | Few accuracy: 18.0 -> 24.9 | Helped |
| Add BatchFormer | PaCo, ResNet-50 · ImageNet-LT | Few accuracy: 39.1 -> 42.1 | Helped |
| Add BatchFormer | RIDE-3e · iNaturalist 2018 | All accuracy: 72.5 -> 74.1 | Helped |
| Add BatchFormer | CGE · UT-Zap50K | AUC: 31.5 -> 34.6 | Helped |
| Add BatchFormer | SWAD · OfficeHome | Avg. accuracy: 62.1 -> 64.3 | Helped |
| Add BatchFormer | SWAD · TerraIncognita | Avg. accuracy: 42.1 -> 44.8 | Helped |
| Add BatchFormer | MoCo-v2 · ImageNet linear classification | Top-1: 67.5 -> 68.4 | Helped |
| Add BatchFormer | MoCo-v3 · ImageNet linear classification | Top-1: 68.9 -> 69.8 | Helped |
| Add BatchFormer | ResNet50 · ImageNet image recognition | Top-1: 78.9 -> 78.9 | No clear effect |
| Remove the shared classifier (ablation) | BatchFormer with Balanced Softmax, ResNet-50 · ImageNet-LT | Few accuracy: 34.1 -> 42.2 | Helped |
Closest papers by meaning
- Decoupling Representation and Classifier for Long-Tailed Recognition (ICLR 2020 · similarity 0.85)
- Long-tailed Recognition with Model Rebalancing (NeurIPS 2025 · similarity 0.85)
- Learning Imbalanced Data With Vision Transformers (CVPR 2023 · similarity 0.85)
- Learning from Neighbors: Category Extrapolation for Long-Tail Learning (CVPR 2025 · similarity 0.84)
- Long-Tailed Recognition via Information-Preservable Two-Stage Learning (NeurIPS 2025 · similarity 0.83)
- DiffuLT: Diffusion for Long-tail Recognition Without External Knowledge (NeurIPS 2024 · similarity 0.83)
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.