aiqbench analysis
This research introduces Experience Blending, a continual learning framework that uses differential-privacy-inspired noise to generate support boundary data, effectively regularizing decision boundaries to mitigate catastrophic forgetting. By jointly training on exemplars and these boundary-adjacent representations, the method achieves significant accuracy improvements across standard benchmarks like CIFAR-10 and ImageNet1K. The approach offers a robust alternative to standard experience replay by enriching the feature space near critical decision thresholds.
Continual Learning with Support Boundary Experience Blending
From arXiv
Full textarXiv:2507.23534v3 Announce Type: replace Abstract: Continual learning (CL) seeks to mitigate catastrophic forgetting when models are trained with sequential tasks. A common approach, experience replay (ER), stores past exemplars but only sparsely approximates the data distribution, yielding fragile and oversimplified decision boundaries. We address this limitation by introducing Support Boundary Data (SBD), generated via differential-privacy-inspired noise into latent features to create boundary-adjacent representations that implicitly regularize decision boundaries. Building on this idea, we propose Experience Blending (EB), a framework that jointly trains on exemplars and SBD through a dual-model aggregation strategy.
EB has two components: (1) latent-space noise injection to generate support boundary data, and (2) end-to-end training that jointly leverages exemplars and SBD. Unlike standard experience replay, SBD enriches the feature space near decision boundaries, leading to more stable and robust continual learning. Extensive experiments on CIFAR-10, CIFAR-100, Tiny ImageNet, and ImageNet1K demonstrate consistent accuracy improvements of 10%, 6%, 13%, 2%, respectively.
Abstract:Continual learning (CL) seeks to mitigate catastrophic forgetting when models are trained with sequential tasks. A common approach, experience replay (ER), stores past exemplars but only sparsely approximates the data distribution, yielding fragile and oversimplified decision boundaries. We address this limitation by introducing Support Boundary Data (SBD), generated via differential-privacy-inspired noise into latent features to create boundary-adjacent representations that implicitly regularize decision boundaries. Building on this idea, we propose Experience Blending (EB), a framework that jointly trains on exemplars and SBD through a dual-model aggregation strategy.
EB has two components: (1) latent-space noise injection to generate support boundary data, and (2) end-to-end training that jointly leverages exemplars and SBD. Unlike standard experience replay, SBD enriches the feature space near decision boundaries, leading to more stable and robust continual learning. Extensive experiments on CIFAR-10, CIFAR-100, Tiny ImageNet, and ImageNet1K demonstrate consistent accuracy improvements of 10%, 6%, 13%, 2%, respectively.
Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2507.23534 [cs.LG]
(or arXiv:2507.23534v3 [cs.LG] for this version)
Full text under the publisher's open licence · 165 words
Related coverage

EvoLib: Turning experience into evolving knowledge
Microsoft Research · Jul 30, 2026
Principles that Guide, Actions that Inform: Agent Evolution via Knowledge Abstraction
arXiv · Oct 7, 2026
REVOLVE: An Automated Closed-Loop Framework for Evolving Robot Manipulation with Minimal Human Intervention
arXiv · Sep 15, 2026
Beyond Episodic Recall: Experience-to-Knowledge Evolution for Lifelong Embodied Navigation
arXiv · Nov 18, 2026