Tag
This research paper introduces a parametric multimodal user memory system that improves AI agents' ability to recall users by integrating perceptual data like voice and appearance, using vision-language models and dedicated encoders to surpass text-based methods.