Research map / Continual learning / Class-incremental
Flexi-FSCIL: Adaptive Knowledge Retention for Breaking the Stability-Plasticity Dilemma in Few-Shot Class-Incremental Learning
Wufei Xie, Yalin Wang, Chenliang Liu, Zhaohui Jiang, Xue Yang
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 AGRF (gated frozen+trainable fusion) to the base model (ablation) | Flexi-FSCIL · CUB200 | accuracy delta vs no strategy: +34.08 | Improved |
| add ADHD distillation on top of AGRF (ablation) | Flexi-FSCIL · CUB200 | last session accuracy (%): 52.77 -> 55.39 | Improved |
| add unlabeled data without POE filtering (ablation) | Flexi-FSCIL · CUB200 | last session accuracy (%): 55.39 -> 58.32 | Improved |
| add POE selection of unlabeled samples (ablation) | Flexi-FSCIL · CUB200 | last session accuracy (%): 58.32 -> 66.24 | Improved |
| add both FA and SP criteria in POE (ablation) | Flexi-FSCIL · CUB200 | last session accuracy (%): 59.51 -> 66.24 | Improved |
| use AvgPool instead of spatial, channel or flatten similarity in ADHD (ablation) | Flexi-FSCIL · CUB200 | PD (performance drop): 14.09 -> 12.97 | Improved |
| increase number of unlabeled samples beyond 160 (to 200) (ablation) | Flexi-FSCIL · CUB200 | last session accuracy (%): 66.24 -> 64.02 | Hurt |
| use POE with 40 unlabeled samples vs without POE (ablation) | Flexi-FSCIL · CUB200 | last session accuracy (%): 56.93 -> 63.19 | Improved |
| embed Flexi-FSCIL into FACL | FACL · CUB200 | PD (performance drop): 18.04 -> 16.84 | Improved |
| embed Flexi-FSCIL into FACT | FACT · CUB200 | PD (performance drop): 21.55 -> 20.32 | Improved |
| Flexi-FSCIL vs best baseline on miniImageNet | Flexi-FSCIL · miniImageNet | delta F: +58.29 vs +58.08 | No clear effect |
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.
- CLOSER — CLOSER: Towards Better Representation Learning for Few-Shot Class-Incremental Learning (ECCV 2024) · better on 1 of 1 shared result
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
- Few-Shot Class-Incremental Learning via Class-Aware Bilateral Distillation (CVPR 2023 · similarity 0.91)
- Few-Shot Hybrid Incremental Learning:Continually Learning under Data Scarcity and Task Uncertainty (CVPR 2026 · similarity 0.90)
- Tripartite Weight-Space Ensemble for Few-Shot Class-Incremental Learning (CVPR 2025 · similarity 0.89)
- Adaptive Decision Boundary for Few-Shot Class-Incremental Learning (AAAI 2025 · similarity 0.89)
- Fine-Grained Knowledge Selection and Restoration for Non-exemplar Class Incremental Learning (AAAI 2024 · similarity 0.87)
- Pre-trained Vision and Language Transformers Are Few-Shot Incremental Learners (CVPR 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.
