Frozen Vision-Language Models Rarely Deploy Visual History for Decision-Making视觉语言行动模型视觉历史利用悖论：模型编码但不依赖过去帧
S 1.7T11 sources1 个来源R7-research
Researchers audited how frozen vision-language-action models handle visual history using layer-resolved linear probing and causal interventions across three VLAs from two architecture families.
The study reveals that despite encoding past-frame content throughout the network, history-specific information is nearly absent, with stored history functioning as a largely redundant copy of the present frame.
History becomes causally important only under heavy frame degradation, and different model architectures deploy history differently—one increasingly relies on history as a fallback under occlusion.