Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Learning to Model the Tail
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 |
|---|---|---|---|
| replace plain fine-tuning with MetaModelNet | plain ResNet152 classifier · SUN-397 | accuracy (%): 48.03 -> 57.34 | Helped |
| MetaModelNet vs best resampling baseline (over-sampling) | over-sampling · SUN-397 | accuracy (%): 52.61 -> 57.34 | Helped |
| use sample-size dependent transform with identity regularization (fixed split) vs single model regression (ablation) | model regression [21] · SUN-397 | accuracy (%): 54.68 -> 56.86 | Helped |
| add recursive head-tail splits (ablation) | MetaModelNet fixed split · SUN-397 | accuracy (%): 56.86 -> 57.34 | Helped |
| MetaModelNet on pre-trained ResNet50 features | plain ResNet50 · SUN-397 | accuracy (%): 46.90 -> 54.99 | Helped |
| fine-tune CNN first, then MetaModelNet (Fix FT) (ablation) | MetaModelNet on pre-trained features · SUN-397 | accuracy (%): 54.99 -> 58.53 | Helped |
| fine-tune features progressively during recursive training (Recur FT) (ablation) | Fix FT + MetaModelNet · SUN-397 | accuracy (%): 58.53 -> 58.74 | Helped |
| fine-tuning features for plain baseline (ablation) | plain ResNet50 · SUN-397 | accuracy (%): 46.90 -> 49.40 | Helped |
| MetaModelNet on ResNet50 fine-tuned outperforms deeper ResNet152 plain/ours | ResNet152 MetaModelNet · SUN-397 | accuracy (%): 57.34 -> 58.74 | Helped |
| MetaModelNet on long-tailed Places-205 from scratch | plain AlexNet · Places-205 | accuracy (%): 23.53 -> 30.71 | Helped |
| MetaModelNet on long-tailed ILSVRC-2012 | plain AlexNet · ILSVRC-2012 | accuracy (%): 68.85 -> 73.46 | Helped |
| passing 1-shot models through MetaModelNet | 1-shot model parameters · SUN-397 | average parameter norm: 0.53 -> 1.36 | Helped |
Closest papers by meaning
- CORAL: Disentangling Latent Representations in Long-Tailed Diffusion (NeurIPS 2025 · similarity 0.87)
- Breaking Long-Tailed Learning Bottlenecks: A Controllable Paradigm with Hypernetwork-Generated Diverse Experts (NeurIPS 2024 · similarity 0.87)
- Decoupled Optimisation for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.86)
- DiffuLT: Diffusion for Long-tail Recognition Without External Knowledge (NeurIPS 2024 · similarity 0.86)
- Long-Tailed Learning as Multi-Objective Optimization (AAAI 2024 · similarity 0.86)
- LT-Soups: Bridging Head and Tail Classes via Subsampled Model Soups (NeurIPS 2025 · similarity 0.86)
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.