AI Paper Atlas

Research map / Continual learning / Class-incremental

A Theoretical Bridge Between Long-Tailed Recognition and Continual Learning

Mahdiyar Molahasani, Michael Greenspan, Ali Etemad

NeurIPS 2026 · Continual learning › Class-incremental · approach: Class-balanced knowledge retention

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Theoretically links long-tailed recognition to continual learning by showing long-tailed training drives parameters into an O(1/sqrt(IF)) neighborhood of the dominant-class solution.

Problem

Understanding how training on long-tailed class distributions relates to continual learning, which has not been previously connected to long-tailed recognition theoretically.

Approach

Shows that training on a long-tailed dataset drives model parameters into an O(1/sqrt(IF)) neighborhood of the dominant-class solution under uniform and exponentially decaying cardinality distributions, and builds on this to bridge LTR and CL.

Key results

Measured effects reported in the paper

What was changedWhereMeasuredEffect
learn Head then Tail sequentially with SGP instead of joint trainingResNet-32 baseline · CIFAR100-LT IF=100accuracy (%): 38.3 -> 50.7Improved
use a stronger CL method (SGP) instead of LwF in CLTR (ablation)CLTR · CIFAR100-LT IF=100accuracy (%): 45.1 -> 50.7Improved
use TPL as CL method in CLTR instead of LwF (ablation)CLTR · CIFAR10-LT IF=100accuracy (%): 76.3 -> 84.7Improved
use TPL as CL method in CLTR instead of EWC (ablation)CLTR · ImageNet-LTtop-1 accuracy (%): 48.9 -> 53.9Improved
apply CLTR (SGP) versus weight-balancing WD+MaxCLTR · CIFAR100-LT IF=100accuracy (%): 53.4 -> 50.7Hurt
use DualPrompt CL on CLIP instead of full fine-tuningCLIP ViT-B/32 · CIFAR100-LT IF=100accuracy (%): 73.4 -> 79.2Improved
use CL method instead of naive sequential training (ablation)CLTR (DualPrompt) · CIFAR100-LT IF=100 (CLIP)accuracy (%): 48.0 -> 79.2Improved
add CL regularization to Head-then-Tail training (ablation)CLTR (LwF/EWC) · CIFAR100-LTupper-bound inequalityImproved
use more partitions (N=4) with stronger CL methods, N=2 with weakerCLTR · LTR benchmarksaccuracy (%)No clear effect
prompt learning with CLIP instead of full fine-tuningCLIP ViT-B/32 · CIFAR100-LT IF=100accuracy (%): 73.4 -> 60.4Hurt

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