Affective Agent: On-Device Personalized Intervention Reasoning for Wearable Systems
Source: https://arxiv.org/abs/2609.12322 · platform: arxiv · authors: Reina Mun, Zishen Wan, Vijay Janapa Reddi · date: 2026-09-11
TL;DR（中文摘要）
可穿戴设备的情感计算已能做状态推断，但「是否/何时/如何干预」的端侧推理仍是难点。Affective Agent 给出三层参考架构（感知/个性化/干预），用亚 1B 参数的小型语言模型结合生理证据、上下文与用户历史，在不确定性下决定干预策略，不依赖云端也不需要按用户重训练，配结构化记忆演化。IEEE Internet Computing 可穿戴专刊（2026 Sep/Oct）录用。作为可穿戴 ≤1B 模型干预推理的架构样本值得跟踪。
Summary (English)
Affective Agent is a three-layer reference architecture (perception, personalization, intervention) for on-device personalized intervention reasoning on wearable-class hardware, combining a sub-1B-parameter LLM with physiological evidence and structured memory evolution, without cloud dependency or per-user retraining. Accepted in IEEE Internet Computing (Sep/Oct 2026).
Abstract
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 intervention) with structured memory evolution, and is evaluated across intervention quality, latency, and energy on wearable-class constraints.
基本信息
| 项 | 值 | |------|------| | 论文 ID | 2609.12322 | | 发表 | 2026-09-11 | | 作者 | Reina Mun, Zishen Wan, Vijay Janapa Reddi | | 备注 | Accepted for publication in IEEE Internet Computing, Special Issue on Wearable Computing (Sep/Oct 2026). 4 figures, 2 tables | | abs | <https://arxiv.org/abs/2609.12322> | | PDF | <https://arxiv.org/pdf/2609.12322v1> |
入库依据（同日 digest 交叉验证）
论文流水线 2026-09-16 备选中等相关；abs 页核实（opencli 429 后 fallback）。