AI Paper Atlas

Research map / Continual learning / Class-incremental

Revisiting Class-Incremental Learning with Pre-Trained Models: Generalizability and Adaptivity are All You Need

Da-Wei Zhou, Zi-Wen Cai, Han-Jia Ye, De‐Chuan Zhan, Ziwei Liu

IJCV 2024 · Continual learning › Class-incremental · approach: Exemplar-efficient class-incremental learning · 245 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
adapt the pre-trained model with Adapter in the first stage and concatenate with frozen PTM featuresSimpleCIL · ImageNet-R B0 Inc5last-stage accuracy: 54.55 -> 64.33Improved
adapt with SSF and concatenate with PTM featuresSimpleCIL · ObjectNet B0 Inc10last-stage accuracy: 53.59 -> 56.64Improved
adapt with Finetune and concatenate with PTM featuresSimpleCIL · ObjectNet B0 Inc10last-stage accuracy: 53.59 -> 48.34Hurt
adapt with Finetune and concatenate with PTM featuresSimpleCIL · OmniBenchmark B0 Inc30last-stage accuracy: 73.15 -> 65.03Hurt
replace sequential finetuning with frozen PTM prototype classifierFinetune · CIFAR100 B0 Inc5last-stage accuracy: 20.17 -> 81.26Improved
use frozen PTM prototype classifier instead of DualPromptDualPrompt · CUB200 B0 Inc10last-stage accuracy: 66.54 -> 86.73Improved
add APER SSF to classical CIL methods setting without exemplarsSimpleCIL · ImageNet-A B0 Inc20last-stage accuracy: 49.44 -> 52.67Improved
adapt in all stages instead of only the first (ablation)APER w/ Adapter · CIFAR100 B0 Inc10accuracyHurt
add adapted model features to PTM features (APER) instead of SimpleCIL-PTM or SimpleCIL-Adapted alone (ablation)SimpleCIL · ImageNet-A Base100 Inc5accuracyImproved
project APER features to 30 dims with PCA (ablation)APER w/ SSF · ObjectNet B100 Inc5accuracyNo clear effect
use APER with domain-incremental SSFSimpleCIL · Office-Homelast-stage accuracy: 75.85 -> 83.14Improved
VPT-Shallow adaptationSimpleCIL · Office-Homelast-stage accuracy: 75.85 -> 63.26Hurt

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.

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