AI Paper Atlas

Research map / Continual learning / Online and replay-based

Quantized Gradient Projection for Memory-Efficient Continual Learning

Dongjun Kim, Seohyeon Cha, Huancheng Chen, Chianing Wang, Haris Vikalo

ICLR 2026 · Continual learning › Online and replay-based · approach: Unified and unsupervised continual frameworks

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
quantize GPM to 8 bitsGPM-FP · four CL benchmarksACC (%): 71.11 -> 70.70No clear effect
quantize GPM to 4 bitsGPM-FP · four CL benchmarksACC (%): 71.11 -> 69.74No clear effect
8QGPM instead of memory-matched rehearsal (DER++-MC)DER++-MC · 10-split CIFAR100ACC (%): 60.63 -> 70.70Improved
reduce quantization bitwidth from 8 to 4 bits (ablation)QGPM · 10-split CIFAR100ACC (%): 65.01 -> 25.03Hurt
add outlier handling (2.0% outliers kept in full precision) (ablation)QGPM · 10-split CIFAR100ACC (%): 41.23 -> 55.36Improved
add quantization-error-aware scaling (alpha=80) to relax orthogonality (ablation)QGPM · 10-split CIFAR100ACC (%): 44.20 -> 67.26Improved
use large alpha (200) instead of moderate alpha (80) (ablation)QGPM · 10-split CIFAR100ACC (%): 67.26 -> 65.17Hurt
add sparse sketching (ablation)8QGPM · miniImageNet ViT-SACC (%): 72.21 -> 72.87No clear effect
add sparse sketching to reduce GPM construction time (ablation)8QGPM · miniImageNet ViT-SGPM construction time (s): 115.06 -> 10.60Improved
add sparse sketching to reduce training time (ablation)8QGPM · miniImageNet ViT-Stotal training time (s): 1768 -> 820Improved
8QGPM on ViT trained from scratchGPM-FP · 10-split miniImageNet (empty ViT)ACC (%): 48.47 -> 48.39No clear effect
8QGPM on NLP text classificationFull precision GPM · Yahoo Answers (5 tasks)ACC (%): 77.80 -> 77.97No clear effect

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