aiqbench analysis
This research addresses the critical gap in personalized sleep guidance by leveraging large language models to reason over unannotated wearable data, bypassing the need for costly expert annotation. The proposed two-stage framework successfully distills complex reasoning into lightweight small language models suitable for local deployment, balancing privacy with high-quality inference that outperforms commercial and open-source alternatives.
Toward Personalized Sleep Guidance from Wearable Data Using Language Models
arXiv:2609.22463v1 Announce Type: new Abstract: Sleep monitoring using wearable data has shown promise for personal health, yet large language model (LLM)-based summarization and question answering remain insufficient for personalized sleep guidance. Training specialized models, however, often requires costly expert annotation. Moreover, privacy and accessibility concerns motivate lightweight, local deployment for end users. We present a two-stage framework to address these challenges. Specifically, in Stage~1, a multi-agent LLM pipeline reasons structured sleep guidance from unannotated wearable records, enabling scalable dataset construction. Stage~2 distills guidance reasoning trajectories into small language models (SLMs) through supervised fine-tuning and integrates a training-free Best-of-$N$ selection strategy to enhance inference. Experimental results demonstrate our method outperforms commercial general and medical LLMs and open-source models. Human evaluation further supports the quality of the generated guidance and the feasibility of personalized sleep guidance with SLMs.
Authors:Yusheng Tan, Running Zhao, Sofia Angel, Ninghui Hao, Ash Arian, Nikita N. Dulin, Jay Lin, Ou Zhu, Faiza Shaik, Xinxing Yang, Bonnie W. Leung, Katie Roster, Arlene Ruiz de Luzuriaga, Kenneth Lee, Alejandra Lastra, Habibul Ahsan, Guihong Wan
Abstract:Sleep monitoring using wearable data has shown promise for personal health, yet large language model (LLM)-based summarization and question answering remain insufficient for personalized sleep guidance. Training specialized models, however, often requires costly expert annotation. Moreover, privacy and accessibility concerns motivate lightweight, local deployment for end users. We present a two-stage framework to address these challenges. Specifically, in Stage~1, a multi-agent LLM pipeline reasons structured sleep guidance from unannotated wearable records, enabling scalable dataset construction. Stage~2 distills guidance reasoning trajectories into small language models (SLMs) through supervised fine-tuning and integrates a training-free Best-of-$N$ selection strategy to enhance inference. Experimental results demonstrate our method outperforms commercial general and medical LLMs and open-source models. Human evaluation further supports the quality of the generated guidance and the feasibility of personalized sleep guidance with SLMs.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2609.22463 [cs.CL]
(or arXiv:2609.22463v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2609.22463 arXiv-issued DOI via DataCite (pending registration)
Full text under the publisher's open licence · 191 words
Related coverage
SensorFM: Towards a general intelligence and interface for wearable health data
Google Research · Jul 9, 2026
The best smartwatches of 2026: Expert tested and reviewed
ZDNet · Jul 29, 2026
2026.8: Approachable by design
Home Assistant · Aug 5, 2026
The wearable health boom is creating a data overload for doctors - what happens next
ZDNet · Jun 13, 2026