Research map / Continual learning / Domain and prompt-based
Hierarchical Decomposition of Prompt-Based Continual Learning: Rethinking Obscured Sub-optimality
Liyuan Wang, Jingyi Xie, Xingxing Zhang, Mingyi Huang, Hang Su, Jun Zhu
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 |
|---|---|---|---|
| replace existing prompt-based continual-learning methods with HiDe-Prompt | ViT-B/16 continual learning · Split CIFAR-100 and Split ImageNet-R across supervised and self-supervised pre-training | FAA: up to 15.01% and 9.61% lead | Improved |
| add prompt ensemble and WTP (ablation) | naive task-specific prompt architecture · Split CIFAR-100 and Split ImageNet-R, across pre-training paradigms | FAA: 72.20 -> 75.93; 53.68 -> 60.86 | Improved |
| add TII to WTP (ablation) | HiDe-Prompt component ablation · Split CIFAR-100, iBOT-1K | FAA: 75.93 -> 78.90 | Improved |
| add TII to WTP (ablation) | HiDe-Prompt component ablation · Split ImageNet-R, iBOT-1K | FAA: 60.86 -> 61.22 | Improved |
| add TAP to WTP (ablation) | HiDe-Prompt component ablation · Split CIFAR-100, iBOT-1K | FAA: 75.93 -> 86.04 | Improved |
| add TAP to WTP (ablation) | HiDe-Prompt component ablation · Split ImageNet-R, iBOT-1K | FAA: 60.86 -> 67.13 | Improved |
| add both TII and TAP to WTP (ablation) | HiDe-Prompt component ablation · Split CIFAR-100, iBOT-1K | FAA: 75.93 -> 90.52 | Improved |
| add both TII and TAP to WTP (ablation) | HiDe-Prompt component ablation · Split ImageNet-R, iBOT-1K | FAA: 60.86 -> 70.66 | Improved |
| add contrastive regularization (ablation) | WTP+TII+TAP · Split CIFAR-100, iBOT-1K | FAA: 90.52 -> 93.48 | Improved |
| add contrastive regularization (ablation) | WTP+TII+TAP · Split ImageNet-R, iBOT-1K | FAA: 70.66 -> 71.33 | Improved |
| increase contrastive-regularization strength (ablation) | WTP with CR · Split CIFAR-100 and Split ImageNet-R, iBOT-1K and DINO-1K | FAA: 75.13 -> 71.53; 61.47 -> 59.52; 74.99 -> 70.87; 57.71 -> 56.37 | Hurt |
| increase contrastive-regularization strength (ablation) | WTP+TII+TAP with CR · Split CIFAR-100 and Split ImageNet-R, iBOT-1K and DINO-1K | FAA: 90.15 -> 93.35; 70.97 -> 71.71; 88.97 -> 92.61; 66.88 -> 68.21 | 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 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
- Mixture of Experts Meets Prompt-Based Continual Learning (NeurIPS 2024 · similarity 0.92)
- Convolutional Prompting meets Language Models for Continual Learning (CVPR 2024 · similarity 0.91)
- Vector Quantization Prompting for Continual Learning (NeurIPS 2024 · similarity 0.91)
- Evolving Parameterized Prompt Memory for Continual Learning (AAAI 2024 · similarity 0.90)
- Consistent Prompting for Rehearsal-Free Continual Learning (CVPR 2024 · similarity 0.90)
- Advancing Prompt-Based Methods for Replay-Independent General Continual Learning (ICLR 2025 · similarity 0.90)
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.
