AI Paper Atlas

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

ECCV 2024 · Continual learning › Domain and prompt-based · approach: Prompt-based continual learning · 10 citations (OpenAlex)

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 changedWhereMeasuredEffect
add second-level semantic residual prompting on top of first-level CLIP keys (ablation)STAR-Prompt · Split ISICfinal average accuracy: 58.87 -> 66.67Improved
add second-level prompts on top of first-level keys (ablation)STAR-Prompt · Split Cars-196final average accuracy: 87.57 -> 87.62No clear effect
learn first-level prompts instead of hand-crafted CLIP templates (ablation)STAR-Prompt · Split Cars-196final average accuracy: 79.88 -> 87.62Improved
use additive semantic residuals instead of prefix tuning (ablation)STAR-Prompt · Split ImageNet-Rfinal average accuracy: 71.34 -> 89.83Improved
use additive semantic residuals instead of prefix tuning (ablation)STAR-Prompt · Split Cars-196final average accuracy: 60.03 -> 87.62Improved
add generative replay (ablation)STAR-Prompt · Split ISICfinal average accuracy: 50.76 -> 66.67Improved
add generative replay (ablation)STAR-Prompt · Split ImageNet-Rfinal average accuracy: 88.55 -> 89.83Improved
use mixture of Gaussians instead of single Gaussian in generative replay (ablation)STAR-Prompt · Split ISICfinal average accuracy: 60.79 -> 66.67Improved
add confidence modulation of residuals (ablation)STAR-Prompt · Split ISICfinal average accuracy: 63.53 -> 66.67Improved
add confidence modulation of residuals (ablation)STAR-Prompt · Split EuroSATfinal average accuracy: 93.68 -> 93.70No clear effect
use STAR-Prompt instead of Zero-shot CLIPCLIP ViT-L/14 · Split EuroSATfinal average accuracy: 54.50 -> 93.70Improved
use STAR-Prompt instead of SLCAcontinual learning on CIFAR-100 · Split CIFAR-100final average accuracy: 91.53 -> 90.12Hurt

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.

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

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy