A January 17, 2026 note on a comparison of architectures for continual learning: wider networks and less global average pooling may favor retention, while ResNets and WideResNets learn new tasks quickly but forget more.
Published on January 17, 2026, the summary reports on a paper comparing architectures for continual learning. According to the publication, wider networks and removing or reducing global average pooling may improve retention. ResNets and WideResNets, in contrast, learn new tasks quickly but show more forgetting.
The note emphasizes that architecture can shift the balance between stability and plasticity, beyond the effect of the continual-learning algorithm. To assess the conclusion, consult the original paper and check which architectures, tasks, metrics, and experimental conditions were compared; the summary does not provide those details.