Research map / Continual learning / Class-incremental
A Tiny Change, A Giant Leap: Long-Tailed Class-Incremental Learning via Geometric Prototype Alignment
Xinyi Lai, Luojun Lin, Weijie Chen, Yuanlong Yu
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 GPA to LUCIR | LUCIR · CIFAR-100-LT shuffled LT-CIL (5 tasks) | average accuracy (%): 35.09 -> 44.68 | Improved |
| add GPA to CODA-Prompt | CODA-Prompt · CIFAR-100-LT shuffled LT-CIL (5 tasks) | average accuracy (%): 65.35 -> 79.20 | Improved |
| add GPA to EASE | EASE · CIFAR-100-LT shuffled LT-CIL (5 tasks) | average accuracy (%): 87.12 -> 89.23 | Improved |
| add GPA to L2P | L2P · CIFAR-100-LT shuffled LT-CIL (5 tasks) | average accuracy (%): 65.83 -> 64.85 | Hurt |
| use GPA instead of GVAlign with LUCIR | LUCIR · CIFAR-100-LT shuffled LT-CIL (5 tasks) | average accuracy (%): 42.80 -> 44.68 | Improved |
| add GPA to DualPrompt on ordered LT-CIL | DualPrompt · CIFAR-100-LT ordered LT-CIL (5 tasks) | average accuracy (%): 54.55 -> 76.65 | Improved |
| add geometric prototype alignment (Phase 2 initialization) (ablation) | GPA · CIFAR-100-LT ablation of GPA | average accuracy (%): 40.12 -> 44.68 | Improved |
| add geometric prototype alignment (Phase 2 initialization) (ablation) | GPA · CIFAR-100-LT ablation of GPA | final task accuracy (%): 29.8 -> 35.4 | Improved |
| add geometric prototype alignment (Phase 2 initialization) (ablation) | GPA · CIFAR-100-LT ablation of GPA | forgetting rate: 15.1 -> 6.94 | Improved |
| add dynamic anchoring (Phase 3) (ablation) | GPA · CIFAR-100-LT ablation of GPA | forgetting rate: 20.6 -> 6.94 | Improved |
| add dynamic anchoring (Phase 3) (ablation) | GPA · CIFAR-100-LT ablation of GPA | average accuracy (%): 42.05 -> 44.68 | Improved |
| add dynamic anchoring (Phase 3) (ablation) | GPA · CIFAR-100-LT ablation of GPA | final task accuracy (%): 32.1 -> 35.4 | Improved |
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.
- DualPrompt — DualPrompt: Complementary Prompting for Rehearsal-Free Continual Learning (ECCV 2022) · 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
- On the Stability-Plasticity Dilemma of Class-Incremental Learning (CVPR 2023 · similarity 0.88)
- Striking a Balance Between Stability and Plasticity for Class-Incremental Learning (ICCV 2021 · similarity 0.88)
- DiAPR: Dimensionally-Allocated Prototype Refinement for Non-Exemplar Class Incremental Learning (AAAI 2026 · similarity 0.87)
- Long-Tail Class Incremental Learning via Independent Sub-prototype Construction (CVPR 2024 · similarity 0.87)
- Gradient Reweighting: Towards Imbalanced Class-Incremental Learning (CVPR 2024 · similarity 0.87)
- Navigating Semantic Drift in Task-Agnostic Class-Incremental Learning (ICML 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.
