# Beyond Class Marginals: Bounding Rehearsal Gaps without Freezing Class Co-occurrence

arXiv: https://arxiv.org/abs/2609.25735 · HTML: https://mlsanity.com/paper/2609.25735/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 20 of its 28 references that are in the map.
- First posted on arXiv 22 Sep 2026.

## Builds on (most cited first, 20 of 20)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Neural Tangent Kernel: Convergence and Generalization in Neural Networks](https://mlsanity.com/paper/1806.07572/) (2018)
- [The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalization](https://mlsanity.com/paper/2006.16241/) (2020)
- [Class-Balanced Loss Based on Effective Number of Samples](https://mlsanity.com/paper/1901.05555/) (2019)
- [Dark Experience for General Continual Learning: a Strong, Simple Baseline](https://mlsanity.com/paper/2004.07211/) (2020)
- [Gradient based sample selection for online continual learning](https://mlsanity.com/paper/1903.08671/) (2019)
- [How to train your ViT? Data, Augmentation, and Regularization in Vision Transformers](https://mlsanity.com/paper/2106.10270/) (2021)
- [Online Continual Learning with Maximally Interfered Retrieval](https://mlsanity.com/paper/1908.04742/) (2019)
- [New Insights on Reducing Abrupt Representation Change in Online Continual Learning](https://mlsanity.com/paper/2104.05025/) (2021)
- [Supervised Contrastive Replay: Revisiting the Nearest Class Mean Classifier in Online Class-Incremental Continual Learning](https://mlsanity.com/paper/2103.13885/) (2021)
- [Rethinking Experience Replay: a Bag of Tricks for Continual Learning](https://mlsanity.com/paper/2010.05595/) (2020)
- [Why Random Reshuffling Beats Stochastic Gradient Descent](https://mlsanity.com/paper/1510.08560/) (2015)
- [Random Shuffling Beats SGD after Finite Epochs](https://mlsanity.com/paper/1806.10077/) (2018)
- [Online Prototype Learning for Online Continual Learning](https://mlsanity.com/paper/2308.00301/) (2023)
- [Information-theoretic Online Memory Selection for Continual Learning](https://mlsanity.com/paper/2204.04763/) (2022)
- [Proximal and Federated Random Reshuffling](https://mlsanity.com/paper/2102.06704/) (2021)
- [Improving Online Continual Learning Performance and Stability with Temporal Ensembles](https://mlsanity.com/paper/2306.16817/) (2023)
- [Learn the Time to Learn: Replay Scheduling in Continual Learning](https://mlsanity.com/paper/2209.08660/) (2022)
- [Watch Your Step: Optimal Retrieval for Continual Learning at Scale](https://mlsanity.com/paper/2404.10758/) (2024)

## Abstract (arXiv)

Class-balanced replay controls class frequency but does not determine the interval between successive replay appearances of a class. We study this interval, the rehearsal gap, separately from the class marginal and class co-occurrence, and introduce randomised-pass replay (RPR), which visits each resident class once per shuffled pass. For a fixed set of C resident classes and replay batch size b less than or equal to C, RPR preserves the balanced time-averaged class marginal and bounds every gap by 2*ceil(C/b)-1; a churn-conditional bound applies while the resident set changes. The scheduler uses no future class information and adds no replay examples or forward passes. In a linear-head ER-ACE diagnostic, joint absence from the incoming and replay batches produces a one-sided classifier-bias gradient. Longer absence episodes are associated with larger negative bias displacement, and removing the incoming-loss mask attenuates the scheduling effect. In the primary ER-ACE experiments, RPR improves final average accuracy by 0.72-1.67 percentage points relative to independent class-balanced retrieval under reservoir storage, with positive effects also observed under balanced storage. Pretrained ViTs show positive effects on the tested LT10 streams with small replay batches, while matched larger-batch controls show no material effect. Fixed-cycle and reused-pass controls change more than one temporal statistic, so the experiments do not isolate rehearsal-gap length from all other forms of temporal dependence. The accuracy effects depend on the learner and operating regime.