[论文] What Should World Models Forget? Stratified Retention for Continual Ad...

论文概要

研究领域: CV
作者: Nishit Anand, Ramani Duraiswami, Dinesh Manocha
发布时间: 2026-10-02
arXiv: 2610.03713

中文摘要

持续学习将模型在先前数据上的性能退化视为失败的证据——这一惯例继承自预测目标平稳的场景，在那些场景中正确标签永远正确。但世界模型不满足这一条件。它们的预测目标是环境，而环境会发生变化：获取时准确的知识之后可能变为错误，丢弃它是必需行为而非缺陷。非平稳真值在概念漂移文献和语言模型的时间事实性中已有充分研究，但尚未针对世界模型进行形式化——世界模型的独特性在于它们同时编码了不可修改的知识。我们主张持续式世界模型需要按不变性时间尺度进行分层保留：将物理规律和物体恒存性等不变量与实例级事实分开——前者绝不可修改，后者应在环境变化时尽快更新。标准遗忘指标无法区分'正确更新了过时知识的世界模型'和'遭遇灾难性遗忘的世界模型'，因此将冻结模型排在最高位；而现有物理推理基准仅评估冻结检查点。我们提出差异化保留方法，联合报告适应流上的不变量回归测试与修正延迟，不做聚合。

原文摘要

Continual learning treats degradation on previously seen data as evidence of failure, a convention inherited from settings with a stationary prediction target, where a correct label remains correct indefinitely. World models do not satisfy this condition. Their prediction target is the environment, which changes, so knowledge that was accurate when acquired may later become false, and discarding it is required behavior rather than a defect. Non-stationary ground truth is well studied in the concept drift literature and in the temporal factuality of language models, but has not been formulated for world models, which are distinctive in that they also encode knowledge that must never be revised. We argue that continual world models require retention stratified by invariance timescale, separa...

自动采集于 2026-10-06

#论文 #arXiv #CV #小凯