Research map / Label-efficient and robust learning / Recognition and long-tail
Echoes of the Past: Boosting Long-tail Recognition via Reflective Learning
Qihao Zhao, Yalun Dai, Shen Lin, Wei Hu, Fan Zhang, Jun Liu
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 Knowledge Review (KR) consistency distillation (ablation) | RIDE · ImageNet-LT | accuracy: 56.3 -> 58.0 | Improved |
| add Knowledge Summary (KS) soft class-correlation labels (ablation) | RIDE · ImageNet-LT | accuracy: 56.3 -> 58.4 | Improved |
| add Knowledge Correction (gradient projection) on top of KR+KS (ablation) | RIDE · ImageNet-LT | accuracy: 58.6 -> 59.0 | Improved |
| add KC on top of KR+KS (ablation) | SADE · iNaturalist 2018 | accuracy: 73.8 -> 74.2 | Improved |
| add Correctly Classified Instances (CCI) filtering to KR (ablation) | BSCE+RL · CIFAR-100-LT IF=100 | top-1 accuracy | Improved |
| replace KL-based KR with direct MSE logit matching (ablation) | BSCE+RL · CIFAR-100-LT IF=100 | top-1 accuracy | Hurt |
| set temperature tau to 2 (ablation) | RL · CIFAR-100-LT and ImageNet-LT | accuracy | Improved |
| add RL to SADE | SADE · ImageNet-LT few-shot | few-shot accuracy: 43.5 -> 47.8 | Improved |
| add RL to Softmax baseline | Softmax · CIFAR-100-LT IF=100 | top-1 accuracy: 41.4 -> 41.9 | Improved |
| add RL to RIDE across backbones | RIDE · ImageNet-LT | overall accuracy: 56.3 -> 59.1 (Swin-T) | Improved |
| KR alone slightly reduces head-class accuracy (ablation) | RL · ImageNet-LT | many-shot accuracy | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- MDCS — MDCS: More Diverse Experts with Consistency Self-distillation for Long-tailed Recognition (ICCV 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- DiffuLT: Diffusion for Long-tail Recognition Without External Knowledge (NeurIPS 2024 · similarity 0.88)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.87)
- Long-Tailed Recognition via Information-Preservable Two-Stage Learning (NeurIPS 2025 · similarity 0.86)
- Decoupled Optimisation for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.86)
- Treasure Hunt: Real-time Targeting of the Long Tail using Training-Time Markers (NeurIPS 2025 · similarity 0.86)
- LTGC: Long-tail Recognition via Leveraging LLMs-driven Generated Content (CVPR 2024 · 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.
