Task leaderboard
Last updated
Task leaderboard
1 reported result · Leaderboard page · Dataset details
| Rank | Method & evaluation | Accuracy↑ Higher is better |
|---|---|---|
| 1 | Addressing Wearable Sleep Tracking Inequity Addressing Wearable Sleep Tracking Inequity: A New Dataset and Novel Methods for a Population with Sleep Disorders Evaluation detailsSame model/protocol; 0.857 ± 0.016 Label set: 2-class sleep/wake Weights: not released Reported as of: 2026-11-09 | 0.857 |
One reported result is available; this is not a multi-method comparison. Rows are ranked only within this task × dataset × metric. Protocols can differ; read each method’s evaluation details before comparing.
From the hub's dated sync history: new results as they are added, and SOTA stars as they are awarded or move.
Rankings apply within each dataset and metric. Open evaluation details for differences in protocols.
| Rank | Method & evaluation | Accuracy↑ Higher is better |
|---|---|---|
| 1 | Addressing Wearable Sleep Tracking Inequity Addressing Wearable Sleep Tracking Inequity: A New Dataset and Novel Methods for a Population with Sleep Disorders Evaluation detailsSame model/protocol; 0.857 ± 0.016 Label set: 2-class sleep/wake Weights: not released Reported as of: 2026-11-09 | 0.857 |
| Rank | Method & evaluation | Area under ROC↑ Higher is better |
|---|---|---|
| 1 | Addressing Wearable Sleep Tracking Inequity Addressing Wearable Sleep Tracking Inequity: A New Dataset and Novel Methods for a Population with Sleep Disorders Evaluation detailsSame model/protocol; 0.926 ± 0.016 Label set: 2-class sleep/wake Weights: not released Reported as of: 2026-11-09 | 0.926 |
| Rank | Method & evaluation | Cohen's kappa (sleep staging agreement). Alias: `kappa` (kept in Rank-axis / axis_id for slug stability; exporter canonicalizes to `cohen_kappa`)↑ Higher is better |
|---|---|---|
| 1 | Addressing Wearable Sleep Tracking Inequity Addressing Wearable Sleep Tracking Inequity: A New Dataset and Novel Methods for a Population with Sleep Disorders Evaluation detailsSame model/protocol; Table 2 0.694 ± 0.025 (abstract rounds to 0.695) Label set: 2-class sleep/wake Weights: not released Reported as of: 2026-11-09 | 0.694 |
| Rank | Method & evaluation | Class F1 for one stated class (e.g. SDB events), not macro↑ Higher is better |
|---|---|---|
| 1 | Addressing Wearable Sleep Tracking Inequity Addressing Wearable Sleep Tracking Inequity: A New Dataset and Novel Methods for a Population with Sleep Disorders Evaluation detailsGPBoost + apnea-severity random effect + LSTM post-processing; participant-level 5-fold CV on 80 participants (16 test/fold); E4 BVP/ACC/EDA/TEMP features; mean ± SD 0.823 ± 0.019 (Table 2) Label set: 2-class, wake = positive class Weights: not released Reported as of: 2026-11-09 | 0.823 |
| Rank | Method & evaluation | F1 score (averaging not stated in the source; e.g. FEEL "average F1"). Neutral public label — never ranked with `macro_f1` or `f1`. Use `macro_f1` / `F1 (macro)` only when the paper states macro averaging↑ Higher is better |
|---|---|---|
| 1 | Self-supervised learning of accelerometer data… Self-supervised learning of accelerometer data provides new insights for sleep and its association with mortality Evaluation details · open weightsExternal validation, 53 PSG nights (Leicester + Pennsylvania), pre-trained SleepNet; subject-wise F1 0.66 ± 0.12 Label set: 2-class sleep/wake Weights: open Reported as of: 2026-11-09 | 0.66 |
| Rank | Method & evaluation | Accuracy↑ Higher is better |
|---|---|---|
| 1 | Cole–Kripke sleep–wake scoring Cole–Kripke sleep–wake scoring (Cole et al. 1992) Evaluation detailsTask Night&Day test set n=363; LSTM seq_len=100 accuracy 88.2 (paper Table 2); Device Algorithm baseline ~76.6; Manual Annotation 86.5 Weights: not released Reported as of: 2026-11-09 | 0.882 |
| Rank | Method & evaluation | Class F1 for one stated class (e.g. SDB events), not macro↑ Higher is better |
|---|---|---|
| 1 | Sadeh sleep–wake scoring Sadeh sleep–wake scoring (Sadeh, Sharkey & Carskadon 1994) Evaluation detailsHeld-out test n=24 participants; RF sleep-wake F1 73.93%; trained on combined clinic PSG studies (abstract n=134); 30 s resolution Reported as of: 2026-11-09 | 0.7393 |
| Rank | Method & evaluation | Specificity at a stated threshold / operating point↑ Higher is better |
|---|---|---|
| 1 | 40 years of actigraphy in sleep medicine and… 40 years of actigraphy in sleep medicine and current state of the art algorithms Evaluation detailsvan Hees 2015 (forearm-angle heuristic); all 28 Newcastle PSG participants, per-subject metrics averaged (Table 1): acc 76.2, sens 83.6, spec 47.5, F1 79.1 (%); ActiGraph agcounts for count algorithms Label set: 2-class, wake specificity Weights: not released Reported as of: 2026-11-09 | 0.475 |
| Rank | Method & evaluation | Specificity at a stated threshold / operating point↑ Higher is better |
|---|---|---|
| 1 | Actigraphy-based sleep estimation in adolescents and adults Actigraphy-based sleep estimation in adolescents and adults: a comparison with polysomnography using two scoring algorithms Evaluation detailsGT3X+ + Cole-Kripke vs in-home PSG, adults n=9 (Table 4); sens 0.96, acc 0.85 Label set: 2-class, wake specificity, epoch-by-epoch Weights: not released Reported as of: 2026-11-09 | 0.35 |
| Rank | Method & evaluation | Accuracy↑ Higher is better |
|---|---|---|
| 1 | Cole–Kripke sleep–wake scoring Cole–Kripke sleep–wake scoring (Cole et al. 1992) Evaluation detailsProspective test on n=21 (optimized on n=20 of total n=41); abstract: correctly distinguished sleep from wakefulness approximately 88% of the time vs PSG Weights: not released Reported as of: 2026-11-09 | 0.88 |
| Rank | Method & evaluation | Accuracy↑ Higher is better |
|---|---|---|
| 1 | DORMI DORMI (Sleepacta sleep/wake; EU MDR Class I; Banfi lightCNNA Sci Rep 2021) Evaluation detailsConcordance / epoch agreement median±MAD 89.32±3.36%; Spec 92.02±3.11%; Sens 89.23±3.46%; F1 90.88±3.04% Weights: not released Reported as of: 2026-11-09 | 0.8932 |
| Rank | Method & evaluation | Cohen's kappa (sleep staging agreement). Alias: `kappa` (kept in Rank-axis / axis_id for slug stability; exporter canonicalizes to `cohen_kappa`)↑ Higher is better |
|---|---|---|
| 1 | DORMI DORMI (Sleepacta sleep/wake; EU MDR Class I; Banfi lightCNNA Sci Rep 2021) Evaluation detailsBinary sleep/wake, 30 s epochs vs AASM visual PSG; lightCNNA on raw triaxial wrist ACC (AX3); LOSO n=81; median±MAD 0.78±0.07 (mean±SD 0.75±0.13); personalized binarization threshold Weights: not released Reported as of: 2026-11-09 | 0.78 |
| Rank | Method & evaluation | Accuracy↑ Higher is better |
|---|---|---|
| 1 | Sadeh sleep–wake scoring Sadeh sleep–wake scoring (Sadeh, Sharkey & Carskadon 1994) Evaluation detailsOverall agreement with PSG for nondominant-wrist algorithm on nondominant data; abstract reports 91–93% on calibration and validation samples (midpoint 0.92 recorded) Weights: not released Reported as of: 2026-11-09 | 0.92 |
| Rank | Method & evaluation | Specificity at a stated threshold / operating point↑ Higher is better |
|---|---|---|
| 1 | Further validation of actigraphy for sleep studiesEvaluation detailsCole-Kripke (1992) algorithm vs lab PSG, n=21 healthy, one night after adaptation night; sensitivity 99%, accuracy 91% (abstract) Label set: 2-class, wake specificity, minute-by-minute Weights: not released Reported as of: 2026-11-09 | 0.34 |
| Rank | Method & evaluation | F1 score (averaging not stated in the source; e.g. FEEL "average F1"). Neutral public label — never ranked with `macro_f1` or `f1`. Use `macro_f1` / `F1 (macro)` only when the paper states macro averaging↑ Higher is better |
|---|---|---|
| 1 | Self-supervised learning of accelerometer data… Self-supervised learning of accelerometer data provides new insights for sleep and its association with mortality Evaluation details · open weightsInternal validation, subject-wise 5-fold CV, 1113 PSG nights (Raine Gen1+Gen2, Newcastle); subject-wise F1 0.75 ± 0.1; averaging not stated in main text Label set: 2-class sleep/wake Weights: open Reported as of: 2026-11-09 | 0.75 |
| Rank | Method & evaluation | other↑ Higher is better |
|---|---|---|
| 1 | HDCZAEvaluation detailsMean C-statistic for SPT-window vs PSG in sleep-clinic patients n=28; healthy good sleepers n=22 mean C=0.83; Whitehall II n=3752 vs diary SPT bias +10.9/+2.9 min (men/women) Weights: not released Reported as of: 2026-11-09 | 0.86 |