AI RESEARCH
One Adapter for All: Towards Unified Representation in Step-Imbalanced Class-Incremental Learning
arXiv CS.LG
•
ArXi:2603.10237v1 Announce Type: cross Class-incremental learning (CIL) aims to acquire new classes over time while retaining prior knowledge, yet most setups and methods assume balanced task streams. In practice, the number of classes per task often varies significantly. We refer to this as step imbalance, where large tasks that contain classes dominate learning and small tasks inject unstable updates. Existing CIL methods assume balanced tasks and therefore treat all tasks uniformly, producing imbalanced updates that degrade overall learning performance.