Learning transferable human physiology from two million hours of sleep with SleepFM-2
arXiv preprint, cs.AI· 2026
Summary
Rahul Thapa, Christopher Sun, William Theodor Lehn-Schiøler, Sophia Claire Kivelson, Umaer Hanif, Hyatt Moore IV, Harrison G. Zhang, Hafsa Ahmed, Marcus Dige, Niels R. Lorenzen, Elisabeth Roxane M.
Key facts
- Subjects
- 282,511 PSG recordings from 26 cohorts (235,865 used for pretraining); counts reported as recordings, not subjects
- License
- CC BY 4.0 on the arXiv posting (paper text). Code repo https://github.com/zou-group/sleepfm-v2-public — repo terms, weight release and upstream data obligations NOT independently checked here. Majority of the pretraining corpus (BioSerenity) is proprietary and unshareable; NSRR/HSP/UK Biobank routes each carry their own DUA. No commercial-eligibility conclusion drawn.
Updated: Sep 27, 2026