[论文] Affective Agent: On-Device Personalized Intervention Reasoning for Wea...

论文概要

研究领域: ML 作者: Reina Mun, Zishen Wan, Vijay Janapa Reddi 发布时间: 2026-09-15 arXiv: 2609.12322

中文摘要

情感计算推动了可穿戴设备的状态推断，但在设备端推理『是否、何时、如何干预』仍然困难。我们提出 Affective Agent——一个面向可穿戴级硬件上不确定性条件下个性化干预推理的三层参考架构。它将一个参数量不足十亿的紧凑语言模型与生理证据、上下文及用户历史结合，决定何时、如何干预，无需依赖云端或逐用户重训练。架构由三个交互层组成（感知、个性化与推理），通过主机管理的结构化记忆演化而非逐用户权重更新来适应个体用户。我们在室内环境质量控制场景中实例化 Affective Agent，并在留出、模拟器生成的纵向场景上评估，场景涵盖生理变异、上下文、信号质量与干预历史。结果表明，在这一合成评估中，记忆驱动的个性化与两遍结构化推理改善了干预决策。通过将决策层搬到设备端，本工作展示了从可穿戴状态推断走向可穿戴级硬件上闭环个性化干预的路径。

原文摘要

Affective computing has advanced wearable state inference, but on-device reasoning about whether, when, and how to intervene remains challenging. We present Affective Agent, a three-layer reference architecture for personalized intervention reasoning under uncertainty on wearable-class hardware. It combines a compact sub-billion-parameter language model with physiological evidence, context, and user history to decide whether, when, and how to intervene, without cloud dependency or per-user retraining. The architecture is organized into three interacting layers (perception, personalization, and reasoning), adapting to individual users through host-managed structured memory evolution rather than per-user weight updates. We instantiate Affective Agent in indoor environmental quality control a...

#论文 #arXiv #ML #小凯