SleepLM: Natural-Language Intelligence for Human Sleep

Summary

SleepLM is a sleep-specific foundation model that enables natural-language intelligence for human sleep — combining a physiological signal encoder for polysomnography (PSG) with a language interface for clinical reasoning, sleep staging, and personalised sleep insight generation. The model is pretrained on multi-cohort PSG data and supports zero-shot and few-shot transfer across sleep staging, apnea detection, and sleep disorder classification tasks. SleepLM enables clinicians and consumers to query their sleep data in natural language (e.g., "explain my sleep architecture", "identify periods of disturbed breathing").