视频记忆是否真正使用了检索内容？记忆特异性的因果审计
原标题：Does Video Memory Use What It Retrieves? A Causal Audit of Memory Specificity
AI 摘要
该论文提出一种「读取时记忆替换」方法，用于区分视频模型的记忆收益与记忆特异性。在冻结的视频世界模型中，无身份信息的控制记忆在 Ego-Exo4D 和 7-Scenes 上几乎恢复了全部收益，TUM 上约恢复部分，支持「表征修复」解释；WorldMem 表现出分级的内容依赖，SAM 2 则表现出强内容依赖，在 DAVIS 和 MOSEv2 上替换正确空间记忆会显著降低指标。
正文节选
Does Video Memory Use What It Retrieves? A Causal Audit of Memory Specificity Abstract Video models increasingly use memory to preserve information over long sequences, with the assumption that gains come from retrieving and using the correct past content. Standard memory ablations test whether memory helps, but not whether the retrieved content is responsible. We test this directly with read-time memory substitution, which replaces the consumed memory value while leaving the rest of the comput