Scholarlink

A single implementation choice in experience replay—whether replay samples are merged with current data or processed separately—can flip experimental outcomes. We show this overlooked detail alone determines whether methods outperform random baselines, with statistically significant differences on CIFAR-10, CIFAR-100, and TinyImageNet. Popular studies and libraries adopt inconsistent defaults, introducing hidden reproducibility gaps in continual learning research. Our results establish this protocol choice as a critical confounder, making fair evaluation of replay methods impossible without explicit reporting.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Merging Versus Separating Replay Samples in Continual Learning

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Andrii Krutsylo

摘要

A single implementation choice in experience replay—whether replay samples are merged with current data or processed separately—can flip experimental outcomes. We show this overlooked detail alone determines whether methods outperform random baselines, with statistically significant differences on CIFAR-10, CIFAR-100, and TinyImageNet. Popular studies and libraries adopt inconsistent defaults, introducing hidden reproducibility gaps in continual learning research. Our results establish this protocol choice as a critical confounder, making fair evaluation of replay methods impossible without explicit reporting.