Research map / Continual learning / Online and replay-based
One-stage Prompt-based Continual Learning
Youngeun Kim, Yuhang Li, Priyadarshini Panda
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 Query-Pool Regularization loss with cosine similarity and softmax (ablation) | OS-Prompt · ImageNet-R 10-task | Average final accuracy A_N (%): 75.00 -> 75.67 | Improved |
| use intermediate token embedding as prompt query (one-stage) instead of separate query ViT (ablation) | CodaPrompt-style PCL · ImageNet-R 5/10/20-task | Average final accuracy A_N (%): 75.45 -> 74.58 (10-task, with CodaPrompt formation) | Hurt |
| apply one-stage framework to L2P prompt formation (ablation) | Deep L2P · ImageNet-R 10-task | Average final accuracy A_N (%): 71.66 -> 73.43 | Improved |
| apply one-stage framework to DualPrompt formation (ablation) | DualPrompt · ImageNet-R 10-task | Average final accuracy A_N (%): 71.32 -> 72.57 | Improved |
| remove query ViT at inference (one-stage) | OS-Prompt · ViT-B/16 inference | GFLOPs relative to prior PCL (%): 100 -> 50.1 | Improved |
| use one-stage design for training | OS-Prompt · relative training complexity vs ER | Relative training cost: 1 -> 2/3 | Improved |
| use unsupervised DINO pretrained weights instead of supervised ImageNet-1k | PCL methods · ImageNet-R 10-task | Average final accuracy A_N (%): 75.45 -> 67.61 (CodaPrompt) | Hurt |
| vary QR loss weight lambda from 1e-5 to 5e-4 (ablation) | OS-Prompt++ · ImageNet-R 5/10/20-task | Average final accuracy A_N (%): 75.63 -> 75.68 (10-task) | No clear effect |
| use CodaPrompt soft-matching formation in one-stage framework (ablation) | OS-Prompt · ImageNet-R 10-task | Average final accuracy A_N (%): 75.45 -> 74.58 | Hurt |
| add QR loss to OS-Prompt for DomainNet (ablation) | OS-Prompt · DomainNet 5-task | Average final accuracy A_N (%): 72.24 -> 73.32 | 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
- Vector Quantization Prompting for Continual Learning (NeurIPS 2024 · similarity 0.90)
- Convolutional Prompting meets Language Models for Continual Learning (CVPR 2024 · similarity 0.90)
- Evolving Parameterized Prompt Memory for Continual Learning (AAAI 2024 · similarity 0.89)
- Visual Prompt Tuning in Null Space for Continual Learning (NeurIPS 2024 · similarity 0.89)
- Mixture of Experts Meets Prompt-Based Continual Learning (NeurIPS 2024 · similarity 0.89)
- Advancing Prompt-Based Methods for Replay-Independent General Continual Learning (ICLR 2025 · similarity 0.89)
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.
