Research map / Label-efficient and robust learning / Recognition and long-tail
Reframing Long-Tailed Learning via Loss Landscape Geometry
Shenghan Chen, Yiming Liu, Yanzhen Wang, Yujia Wang, Xiankai Lu
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 GKP module (ablation) | CE baseline · CIFAR100-LT r=100 | top-1 accuracy: 38.4 -> 42.1 | Improved |
| add GSA module on top of GKP (ablation) | CE baseline · CIFAR100-LT r=100 | top-1 accuracy: 42.1 -> 42.8 | Improved |
| add GKP module (ablation) | BCL · CIFAR100-LT r=100 | top-1 accuracy: 51.9 -> 52.4 | Improved |
| add GSA module (ablation) | BCL · CIFAR100-LT r=100 | top-1 accuracy: 51.9 -> 52.7 | Improved |
| add both GKP and GSA (ablation) | BCL · CIFAR100-LT r=100 | top-1 accuracy: 51.9 -> 53.2 | Improved |
| use group-specific gradient instead of standard SAM (ablation) | GSA · CIFAR100-LT r=100 | top-1 accuracy: 52.1 -> 53.2 | Improved |
| perturb along head-dominated projected gradient only (GSA-proj) (ablation) | GSA · CIFAR100-LT r=100 | top-1 accuracy: 53.2 -> 46.4 | Hurt |
| set group number G to 4 (ablation) | GKP/GSA grouping · CIFAR10-LT/CIFAR100-LT r=100 | top-1 accuracy | Mixed |
| add GKP+GSA to BCL (ablation) | BCL · CIFAR100-LT Many classes | top-1 accuracy: 67.2 -> 67.3 | No clear effect |
| add GKP+GSA to BCL (ablation) | BCL · CIFAR100-LT Few classes | top-1 accuracy: 32.9 -> 34.9 | Improved |
Closest papers by meaning
- SSE-SAM: Balancing Head and Tail Classes Gradually Through Stage-Wise SAM (AAAI 2025 · similarity 0.87)
- Breaking Long-Tailed Learning Bottlenecks: A Controllable Paradigm with Hypernetwork-Generated Diverse Experts (NeurIPS 2024 · similarity 0.86)
- Long-Tail Class Incremental Learning via Independent Sub-prototype Construction (CVPR 2024 · similarity 0.86)
- Focal-SAM: Focal Sharpness-Aware Minimization for Long-Tailed Classification (ICML 2025 · similarity 0.86)
- Learning Representations on the Unit Sphere: Investigating Angular Gaussian and Von Mises-Fisher Distributions for Online Continual Learning (AAAI 2024 · similarity 0.85)
- Instance-Conditional Timescales of Decay for Non-Stationary Learning (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.
