Research map / Continual learning / Class-incremental
Revisiting Pool-based Prompt Learning for Few-shot Class-incremental Learning
Yongwei Jiang, Yixiong Zou, Yuhua Li, Ruixuan Li
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 basic VPT prompts to vanilla ViT (ablation) | ViT · CUB-200 FSCIL | overall accuracy: 61.66 -> 79.26 | Improved |
| add single local spatial prompt (LSP*) (ablation) | VPT · CUB-200 FSCIL | overall accuracy: 79.26 -> 79.49 | Improved |
| use a local spatial prompt pool instead of a single prompt (ablation) | VPT+LSP · iNF200 FSCIL | overall accuracy: 52.94 -> 56.87 | Improved |
| add global frequency-domain spatial prompting to local prompts (ablation) | LGSP-Prompt · FGVCAircraft FSCIL | overall accuracy: 21.58 -> 25.82 | Improved |
| combine LSP and GSP vs LSP only (ablation) | LGSP-Prompt · FGVCAircraft base session | base accuracy: 28.27 -> 27.67 | Hurt |
| use larger kernel sizes (>11) in LSP (ablation) | LSP · FSCIL | accuracy | Hurt |
| vary frequency ring count in GSP (ablation) | GSP · FSCIL | accuracy | No clear effect |
| increase local prompt pool size (ablation) | LSP · FSCIL | accuracy | Improved |
| use prompt pools (CODA-Prompt) vs non-prompt methods in FSCIL | CODA-Prompt · CUB-200 FSCIL | average accuracy: CODA-Prompt 74.63 vs TEEN 78.11 | Hurt |
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
- SEC-Prompt:SEmantic Complementary Prompting for Few-Shot Class-Incremental Learning (CVPR 2025 · similarity 0.89)
- Learning Conditional Space-Time Prompt Distributions for Video Class-Incremental Learning (CVPR 2025 · similarity 0.87)
- SA²VP: Spatially Aligned-and-Adapted Visual Prompt (AAAI 2024 · similarity 0.86)
- Pre-trained Vision and Language Transformers Are Few-Shot Incremental Learners (CVPR 2024 · similarity 0.86)
- CoPL: Contextual Prompt Learning for Vision-Language Understanding (AAAI 2024 · similarity 0.85)
- Understanding prompt engineering may not require rethinking generalization (ICLR 2024 · similarity 0.85)
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.
