Research map / Continual learning / Domain and prompt-based
Semantic Residual Prompts for Continual Learning
Martin Menabue, Emanuele Frascaroli, Matteo Boschini, Enver Sangineto, Lorenzo Bonicelli, Angelo Porrello, SIMONE CALDERARA
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 second-level semantic residual prompting on top of first-level CLIP keys (ablation) | STAR-Prompt · Split ISIC | final average accuracy: 58.87 -> 66.67 | Improved |
| add second-level prompts on top of first-level keys (ablation) | STAR-Prompt · Split Cars-196 | final average accuracy: 87.57 -> 87.62 | No clear effect |
| learn first-level prompts instead of hand-crafted CLIP templates (ablation) | STAR-Prompt · Split Cars-196 | final average accuracy: 79.88 -> 87.62 | Improved |
| use additive semantic residuals instead of prefix tuning (ablation) | STAR-Prompt · Split ImageNet-R | final average accuracy: 71.34 -> 89.83 | Improved |
| use additive semantic residuals instead of prefix tuning (ablation) | STAR-Prompt · Split Cars-196 | final average accuracy: 60.03 -> 87.62 | Improved |
| add generative replay (ablation) | STAR-Prompt · Split ISIC | final average accuracy: 50.76 -> 66.67 | Improved |
| add generative replay (ablation) | STAR-Prompt · Split ImageNet-R | final average accuracy: 88.55 -> 89.83 | Improved |
| use mixture of Gaussians instead of single Gaussian in generative replay (ablation) | STAR-Prompt · Split ISIC | final average accuracy: 60.79 -> 66.67 | Improved |
| add confidence modulation of residuals (ablation) | STAR-Prompt · Split ISIC | final average accuracy: 63.53 -> 66.67 | Improved |
| add confidence modulation of residuals (ablation) | STAR-Prompt · Split EuroSAT | final average accuracy: 93.68 -> 93.70 | No clear effect |
| use STAR-Prompt instead of Zero-shot CLIP | CLIP ViT-L/14 · Split EuroSAT | final average accuracy: 54.50 -> 93.70 | Improved |
| use STAR-Prompt instead of SLCA | continual learning on CIFAR-100 · Split CIFAR-100 | final average accuracy: 91.53 -> 90.12 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- SLCA — SLCA: Slow Learner with Classifier Alignment for Continual Learning on a Pre-trained Model (ICCV 2023) · better on 3 of 5 shared results
- AttriCLIP — AttriCLIP: A Non-Incremental Learner for Incremental Knowledge Learning (CVPR 2023) · better on 2 of 2 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
- Convolutional Prompting meets Language Models for Continual Learning (CVPR 2024 · similarity 0.94)
- Dynamic Cross-Modal Prompt Generation for Multimodal Continual Instruction Tuning (NeurIPS 2026 · similarity 0.91)
- One Size Fits All for Semantic Shifts: Adaptive Prompt Tuning for Continual Learning (ICML 2024 · similarity 0.90)
- Evolving Parameterized Prompt Memory for Continual Learning (AAAI 2024 · similarity 0.90)
- REP: Resource-Efficient Prompting for Rehearsal-Free Continual Learning (NeurIPS 2025 · similarity 0.90)
- Vector Quantization Prompting for Continual Learning (NeurIPS 2024 · 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.
