Research map / Label-efficient and robust learning / Recognition and long-tail
Long-tailed Recognition by Routing Diverse Distribution-Aware Experts
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 |
|---|---|---|---|
| Use RIDE with four experts instead of LFME | Long-tailed recognition with ResNet-32 · CIFAR100-LT | accuracy (All): 43.8 -> 49.1 | Helped |
| Use RIDE with two experts instead of LWS | Long-tailed recognition with ResNet-50 · ImageNet-LT | top-1 accuracy (%): 47.7 -> 54.4 | Helped |
| Use RIDE with four experts instead of BBN | Long-tailed recognition with ResNet-50 · iNaturalist 2018 | accuracy (All): 66.3 -> 72.6 | Helped |
| Use RIDE with three experts instead of LDAM-DRW | Swin-T · ImageNet-LT | top-1 accuracy (Overall): 50.6 -> 54.2 | Helped |
| Use two experts rather than one baseline expert (ablation) | RIDE with LDAM on ResNet-32 · CIFAR100-LT | accuracy: 42.0 -> 44.7 | Helped |
| Increase the expert count from two to four before adding training components (ablation) | RIDE with LDAM on ResNet-32 · CIFAR100-LT | accuracy: 44.7 -> 46.3 | Helped |
| Add individual classification loss instead of collaborative loss (ablation) | Four-expert RIDE with LDAM · CIFAR100-LT | accuracy: 46.3 -> 47.8 | Helped |
| Add distribution-aware diversity loss (ablation) | Four-expert RIDE with individual classification loss · CIFAR100-LT | accuracy: 47.8 -> 48.7 | Helped |
| Add self-distillation from a six-expert model (ablation) | Four-expert RIDE with diversity loss · CIFAR100-LT | accuracy: 48.7 -> 49.3 | Helped |
| Add dynamic expert routing (ablation) | Four-expert RIDE with distillation · CIFAR100-LT | accuracy: 49.3 -> 49.1 | Hurt |
| Add dynamic expert routing (ablation) | Four-expert RIDE with distillation · CIFAR100-LT | GFlops: 1.8x -> 1.3x | Helped |
| Distill a two-expert student from six experts using KD (ablation) | RIDE student · CIFAR100-LT | accuracy: 46.6 -> 47.3 | Helped |
Closest papers by meaning
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.89)
- Long-Tailed Recognition via Information-Preservable Two-Stage Learning (NeurIPS 2025 · similarity 0.88)
- Long-tailed Recognition with Model Rebalancing (NeurIPS 2025 · similarity 0.88)
- Breaking Long-Tailed Learning Bottlenecks: A Controllable Paradigm with Hypernetwork-Generated Diverse Experts (NeurIPS 2024 · similarity 0.88)
- DiffuLT: Diffusion for Long-tail Recognition Without External Knowledge (NeurIPS 2024 · similarity 0.87)
- Long-Tailed Learning as Multi-Objective Optimization (AAAI 2024 · similarity 0.87)
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.