OpenMHC Benchmark

OpenMHC Leaderboard

Wearable & mobile health benchmark on MyHeartCounts. Track 1 (predictive tasks) predicts weekly health outcomes — demographics, medical risk, vitals, mental well-being, and lifestyle — from 168-hour sensor embeddings. The generative tasks operate on the raw signals: Track 2a (imputation) reconstructs masked daily, minute-level signals, and Track 2b (forecasting) predicts future hourly signals. Each method is ranked by skill score vs a track baseline, computed live from the per-user evaluation substrate.

📤 Submit a model ⚙️ Code 📊 Dataset · coming soon 📄 Paper · coming soon 🏠 MyHeartCounts 🤗 Models

Cohort

Track 1 · Predictive Tasks

All participants · 1,762 users evaluated

#	Method	Type ▾ Statistical Models Supervised Neural Models Time-Series Foundation Models Wearable Foundation Models	R ↓	S ↑	S_worst ↑	Demo. ↑	Medical ↑	Vitals ↑	Mental ↑	Lifestyle ↑	Fallback ↓	Submitter
1	LSM-2	Wearable Foundation Models	2.2	+15.1	+13.4	+41.1	-1.8	+22.2	+0.7	+13.4	0.0	OpenMHC team
2	XGBoost	Statistical Models	3.5	+11.6	+8.3	+46.9	-1.4	+13.6	-8.5	+7.3	0.0	OpenMHC team
3	MultiRocket	Statistical Models	3.7	+7.1	+4.6	+27.7	-4.5	+5.6	+0.2	+6.6	0.0	OpenMHC team
4	Linear (baseline)	Statistical Models	4.4	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
5	WBM	Wearable Foundation Models	4.8	+3.3	-0.1	+22.2	-2.0	+6.6	-7.5	-2.8	62.8	OpenMHC team
6	GRU-D	Supervised Neural Models	4.9	+3.1	+0.0	+18.2	-6.2	+10.5	-3.9	-3.3	0.0	OpenMHC team
7	Chronos-2	Time-Series Foundation Models	6.0	-3.4	-5.6	+6.2	-9.6	-7.1	-7.6	+0.9	0.0	OpenMHC team
8	Toto	Time-Series Foundation Models	6.5	-5.0	-7.8	-0.9	-7.6	-4.5	-9.2	-2.6	0.0	OpenMHC team

Age < 40 · 842 users evaluated

#	Method	Type ▾ Statistical Models Supervised Neural Models Time-Series Foundation Models Wearable Foundation Models	R ↓	S ↑	S_worst ↑	Demo. ↑	Medical ↑	Vitals ↑	Mental ↑	Lifestyle ↑	Fallback ↓	Submitter
1	LSM-2	Wearable Foundation Models	2.3	+14.3	+13.4	+23.4	+5.4	+21.5	+4.2	+17.0	0.0	OpenMHC team
2	XGBoost	Statistical Models	3.6	+8.3	+8.3	+21.9	+4.5	+11.3	-3.4	+7.3	0.0	OpenMHC team
3	MultiRocket	Statistical Models	3.7	+7.6	+4.6	+11.1	+0.7	+7.8	+8.2	+10.2	0.0	OpenMHC team
4	WBM	Wearable Foundation Models	4.5	+4.7	-0.1	+12.5	+2.3	+8.8	-0.9	+0.7	62.8	OpenMHC team
5	GRU-D	Supervised Neural Models	5.2	+3.0	+0.0	+8.0	+0.1	+7.4	-1.4	+1.1	0.0	OpenMHC team
6	Chronos-2	Time-Series Foundation Models	5.3	-2.7	-5.6	-7.5	+1.3	-9.2	-3.3	+5.4	0.0	OpenMHC team
7	Linear (baseline)	Statistical Models	5.6	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
8	Toto	Time-Series Foundation Models	5.7	-2.8	-7.8	-13.7	+0.4	-0.9	-2.5	+2.8	0.0	OpenMHC team

Age 40+ · 888 users evaluated

#	Method	Type ▾ Statistical Models Supervised Neural Models Time-Series Foundation Models Wearable Foundation Models	R ↓	S ↑	S_worst ↑	Demo. ↑	Medical ↑	Vitals ↑	Mental ↑	Lifestyle ↑	Fallback ↓	Submitter
1	LSM-2	Wearable Foundation Models	2.2	+14.0	+13.4	+40.7	-0.5	+23.2	+2.5	+4.1	0.0	OpenMHC team
2	XGBoost	Statistical Models	3.5	+11.6	+8.3	+49.9	-0.7	+15.5	-7.5	+0.8	0.0	OpenMHC team
3	MultiRocket	Statistical Models	4.3	+5.7	+4.6	+32.6	-2.4	+5.4	-2.8	-4.3	0.0	OpenMHC team
4	GRU-D	Supervised Neural Models	4.4	+2.8	+0.0	+20.4	-4.4	+13.5	+0.3	-16.1	0.0	OpenMHC team
5	Linear (baseline)	Statistical Models	4.4	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
6	WBM	Wearable Foundation Models	5.0	-0.1	-0.1	+23.3	-1.3	+4.9	-10.9	-16.5	62.8	OpenMHC team
7	Chronos-2	Time-Series Foundation Models	5.7	-5.6	-5.6	+5.0	-10.1	-5.5	-6.0	-11.2	0.0	OpenMHC team
8	Toto	Time-Series Foundation Models	6.4	-7.8	-7.8	-0.6	-7.7	-5.3	-9.3	-16.3	0.0	OpenMHC team

Female · 343 users evaluated

#	Method	Type ▾ Statistical Models Supervised Neural Models Time-Series Foundation Models Wearable Foundation Models	R ↓	S ↑	S_worst ↑	Demo. ↑	Medical ↑	Vitals ↑	Mental ↑	Lifestyle ↑	Fallback ↓	Submitter
1	LSM-2	Wearable Foundation Models	2.4	+20.6	+13.4	+54.5	+4.9	+24.5	+0.1	+18.9	0.0	OpenMHC team
2	XGBoost	Statistical Models	3.3	+14.6	+8.3	+56.3	+3.4	+18.1	-16.9	+12.1	0.0	OpenMHC team
3	MultiRocket	Statistical Models	4.0	+12.3	+4.6	+39.9	+10.3	+8.8	-6.3	+8.6	0.0	OpenMHC team
4	GRU-D	Supervised Neural Models	4.7	+5.2	+0.0	+19.1	+2.7	+12.7	-5.8	-2.8	0.0	OpenMHC team
5	Linear (baseline)	Statistical Models	4.9	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
6	WBM	Wearable Foundation Models	5.5	+1.5	-0.1	+22.8	-3.6	+8.7	-18.5	-1.9	62.8	OpenMHC team
7	Toto	Time-Series Foundation Models	5.5	-2.1	-7.8	+26.2	-7.9	-14.7	-17.8	+3.9	0.0	OpenMHC team
8	Chronos-2	Time-Series Foundation Models	5.8	-3.5	-5.6	+23.8	-9.6	-19.7	-18.6	+6.7	0.0	OpenMHC team

Male · 1,331 users evaluated

#	Method	Type ▾ Statistical Models Supervised Neural Models Time-Series Foundation Models Wearable Foundation Models	R ↓	S ↑	S_worst ↑	Demo. ↑	Medical ↑	Vitals ↑	Mental ↑	Lifestyle ↑	Fallback ↓	Submitter
1	LSM-2	Wearable Foundation Models	2.2	+13.4	+13.4	+32.9	-3.4	+22.5	+1.0	+14.1	0.0	OpenMHC team
2	XGBoost	Statistical Models	3.5	+10.0	+8.3	+37.3	-3.9	+13.9	-6.1	+8.6	0.0	OpenMHC team
3	MultiRocket	Statistical Models	4.1	+4.6	+4.6	+14.6	-7.2	+6.3	+0.7	+8.4	0.0	OpenMHC team
4	WBM	Wearable Foundation Models	4.5	+2.9	-0.1	+18.8	-2.5	+5.9	-5.7	-1.9	62.8	OpenMHC team
5	Linear (baseline)	Statistical Models	5.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
6	GRU-D	Supervised Neural Models	5.2	+0.0	+0.0	+4.1	-7.9	+9.0	-2.4	-2.6	0.0	OpenMHC team
7	Toto	Time-Series Foundation Models	5.7	-0.4	-7.8	+14.8	-8.2	+0.8	-7.8	-1.4	0.0	OpenMHC team
8	Chronos-2	Time-Series Foundation Models	5.7	-0.5	-5.6	+12.8	-10.4	-1.8	-5.1	+2.0	0.0	OpenMHC team

Metric legend — Track 1 predicts weekly health outcomes from 168-hour sensor embeddings; scores are computed vs the Linear baseline.
R (Average Rank) — mean cross-method rank across the outcome tasks; 1 = best (lower is better).
S (Skill Score) — category-balanced % improvement over Linear across tasks (per-task AUPRC / Spearman / Pearson, paired-bootstrap mean); higher is better.
S_worst (Worst-subgroup skill) — the method's lowest skill score across the four demographic cohorts (age <40 / 40+, female / male): its performance for the worst-served subgroup; higher is better.
Demo. / Medical / Vitals / Mental / Lifestyle — per-category skill on that outcome group's tasks (Demographics; Medical Conditions & Risk; Vitals & Blood Biomarkers; Mental Well-Being; Sleep & Lifestyle), category-balanced like S.
Fallback — % of test predictions substituted by the Linear baseline when the method produced no valid output (lower is better).
Source: MyHeartCounts/OpenMHC-leaderboard-data.
Cohort — the selector above recomputes R, S, and the per-category skills within one demographic cohort (a near-median age split at 40, or sex); the count above each table is the cohort's evaluated users. S_worst is the minimum cohort skill across the four cohorts, so it reads the same in every view. Cohorts are smaller than the full population, so scores are noisier; tasks without enough users (or degenerate within the cohort) drop from the aggregates.

Track 2a · Imputation

All participants · 1,777 users evaluated

#	Method	Type ▾ Neural Statistical	R ↓	S ↑	S_worst ↑	Activity ↑	Physio. ↑	Sleep ↑	Workout ↑	Semantic ↑	Fallback ↓	Submitter
Single-day imputation
1	LSM-2 (daily)	Neural	3.8	+61.4	+59.4	+40.0	+31.4	+90.0	+94.9	+30.2	0.0	OpenMHC team
2	Linear	Statistical	7.0	+21.5	+20.0	+4.5	+9.8	+62.6	+56.5	-0.8	0.0	OpenMHC team
3	BRITS	Neural	7.8	+6.8	+0.4	+18.8	-28.5	+39.0	+28.0	-5.7	0.0	OpenMHC team
4	DLinear	Neural	8.2	-5.7	-11.5	+29.3	-5.1	-11.1	+58.2	-45.9	0.0	OpenMHC team
5	LOCF (baseline)	Statistical	8.4	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
6	Temporal mode	Statistical	10.0	-6.2	-12.2	+46.5	-0.7	-13.9	-69.4	-11.8	0.0	OpenMHC team
7	Temporal mean	Statistical	10.4	-31.2	-41.7	-30.7	-18.4	+59.9	-2.1	-93.0	0.0	OpenMHC team
8	Mode	Statistical	10.6	-27.3	-31.5	+46.5	-0.8	-380.7	-69.4	-12.0	0.0	OpenMHC team
9	TimesNet	Neural	10.9	-66.0	-79.1	+9.6	-18.6	-216.2	+0.4	-103.2	0.0	OpenMHC team
10	FEDformer	Neural	11.3	-53.7	-64.1	+28.9	-14.6	-214.6	-53.7	-67.7	0.0	OpenMHC team
11	Mean	Statistical	13.4	-119.7	-133.1	-36.3	-25.5	-380.7	-69.4	-149.8	0.0	OpenMHC team
Long-context imputation (≥ 7×1440 time steps)
1	LSM-2-Sparse (7-day)	Neural	3.3	+64.7	+63.1	+41.0	+34.6	+92.2	+95.7	+34.6	0.0	OpenMHC team
2	LSM-2 (7-day)	Neural	5.5	+46.9	+44.4	+16.7	+19.0	+85.6	+90.6	+8.7	0.0	OpenMHC team
3	Pers. temp. mean	Statistical	9.1	-7.7	-13.5	+1.1	-5.9	+58.9	+15.7	-49.5	0.0	OpenMHC team
4	DLinear (7-day)	Neural	9.5	-28.3	-36.7	+19.9	-2.7	-40.0	+22.9	-69.5	0.0	OpenMHC team
5	Pers. mode	Statistical	10.5	-26.1	-29.4	+46.6	+1.8	-383.1	-69.4	-10.5	0.0	OpenMHC team
6	Pers. mean	Statistical	13.3	-114.1	-120.8	-4.1	-12.6	-437.7	-140.0	-132.5	0.0	OpenMHC team

Age < 40 · 917 users evaluated

#	Method	Type ▾ Neural Statistical	R ↓	S ↑	S_worst ↑	Activity ↑	Physio. ↑	Sleep ↑	Workout ↑	Semantic ↑	Fallback ↓	Submitter
Single-day imputation
1	LSM-2 (daily)	Neural	3.9	+60.8	+59.4	+39.8	+31.2	+90.1	+95.2	+27.2	0.0	OpenMHC team
2	Linear	Statistical	6.9	+21.7	+20.0	+4.0	+9.4	+62.2	+59.2	-1.4	0.0	OpenMHC team
3	BRITS	Neural	7.7	+7.6	+0.4	+21.1	-24.9	+40.4	+32.8	-7.7	0.0	OpenMHC team
4	DLinear	Neural	8.1	-5.4	-11.5	+29.5	-3.2	-15.4	+64.1	-50.0	0.0	OpenMHC team
5	LOCF (baseline)	Statistical	8.3	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
6	Temporal mode	Statistical	10.1	-6.5	-12.2	+46.5	+1.1	-21.4	-60.3	-12.6	0.0	OpenMHC team
7	Temporal mean	Statistical	10.5	-35.2	-41.7	-29.3	-17.0	+58.8	-0.9	-105.3	0.0	OpenMHC team
8	Mode	Statistical	10.6	-27.3	-31.5	+46.5	+1.6	-410.5	-60.3	-12.5	0.0	OpenMHC team
9	TimesNet	Neural	10.8	-66.5	-79.1	+10.6	-16.9	-216.9	+12.4	-112.3	0.0	OpenMHC team
10	FEDformer	Neural	11.2	-56.1	-64.1	+29.0	-12.9	-233.6	-43.3	-74.2	0.0	OpenMHC team
11	Mean	Statistical	13.4	-123.8	-133.1	-34.6	-23.7	-410.5	-60.3	-160.6	0.0	OpenMHC team
Long-context imputation (≥ 7×1440 time steps)
1	LSM-2-Sparse (7-day)	Neural	3.4	+64.0	+63.1	+40.3	+33.6	+91.9	+95.9	+31.9	0.0	OpenMHC team
2	LSM-2 (7-day)	Neural	5.6	+45.5	+44.4	+14.9	+18.1	+86.0	+90.9	+3.4	0.0	OpenMHC team
3	DLinear (7-day)	Neural	9.4	-28.8	-36.7	+20.3	-1.8	-45.1	+30.6	-74.6	0.0	OpenMHC team
4	Pers. temp. mean	Statistical	9.4	-13.5	-13.5	-3.5	-8.4	+56.0	+16.8	-61.0	0.0	OpenMHC team
5	Pers. mode	Statistical	10.5	-25.7	-29.4	+46.5	+3.7	-413.2	-60.3	-10.1	0.0	OpenMHC team
6	Pers. mean	Statistical	13.3	-119.1	-120.8	-8.0	-13.4	-467.4	-119.1	-143.1	0.0	OpenMHC team

Age 40+ · 922 users evaluated

#	Method	Type ▾ Neural Statistical	R ↓	S ↑	S_worst ↑	Activity ↑	Physio. ↑	Sleep ↑	Workout ↑	Semantic ↑	Fallback ↓	Submitter
Single-day imputation
1	LSM-2 (daily)	Neural	3.8	+61.4	+59.4	+40.0	+31.7	+89.9	+94.7	+31.0	0.0	OpenMHC team
2	Linear	Statistical	7.0	+21.5	+20.0	+5.1	+10.3	+62.8	+55.3	-0.5	0.0	OpenMHC team
3	BRITS	Neural	7.9	+5.6	+0.4	+16.3	-32.2	+37.7	+26.9	-5.8	0.0	OpenMHC team
4	DLinear	Neural	8.2	-6.5	-11.5	+28.9	-7.1	-8.5	+55.0	-45.3	0.0	OpenMHC team
5	LOCF (baseline)	Statistical	8.5	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
6	Temporal mode	Statistical	10.0	-6.8	-12.2	+46.3	-3.2	-9.4	-74.2	-12.5	0.0	OpenMHC team
7	Temporal mean	Statistical	10.3	-29.5	-41.7	-32.1	-19.9	+61.0	-1.2	-88.5	0.0	OpenMHC team
8	Mode	Statistical	10.6	-28.1	-31.5	+46.3	-3.6	-364.6	-74.2	-12.7	0.0	OpenMHC team
9	TimesNet	Neural	11.0	-66.9	-79.1	+8.6	-20.6	-214.0	-6.3	-101.2	0.0	OpenMHC team
10	FEDformer	Neural	11.3	-53.6	-64.1	+28.5	-16.6	-204.3	-59.9	-66.3	0.0	OpenMHC team
11	Mean	Statistical	13.5	-118.9	-133.1	-38.1	-27.5	-364.6	-74.2	-146.6	0.0	OpenMHC team
Long-context imputation (≥ 7×1440 time steps)
1	LSM-2-Sparse (7-day)	Neural	3.2	+64.9	+63.1	+41.3	+35.6	+92.3	+95.5	+35.1	0.0	OpenMHC team
2	LSM-2 (7-day)	Neural	5.4	+47.5	+44.4	+18.2	+20.1	+85.4	+90.4	+10.8	0.0	OpenMHC team
3	Pers. temp. mean	Statistical	8.9	-4.3	-13.5	+4.0	-3.9	+61.7	+18.0	-45.4	0.0	OpenMHC team
4	DLinear (7-day)	Neural	9.5	-28.7	-36.7	+19.5	-4.0	-36.9	+18.8	-68.8	0.0	OpenMHC team
5	Pers. mode	Statistical	10.6	-27.2	-29.4	+46.3	-0.4	-366.8	-74.2	-11.8	0.0	OpenMHC team
6	Pers. mean	Statistical	13.4	-112.2	-120.8	-1.8	-12.2	-420.2	-147.6	-129.9	0.0	OpenMHC team

Female · 343 users evaluated

#	Method	Type ▾ Neural Statistical	R ↓	S ↑	S_worst ↑	Activity ↑	Physio. ↑	Sleep ↑	Workout ↑	Semantic ↑	Fallback ↓	Submitter
Single-day imputation
1	LSM-2 (daily)	Neural	3.9	+59.4	+59.4	+38.4	+31.5	+89.4	+94.7	+25.2	0.0	OpenMHC team
2	Linear	Statistical	7.0	+20.0	+20.0	+5.6	+10.8	+63.6	+55.2	-5.1	0.0	OpenMHC team
3	BRITS	Neural	7.9	+0.4	+0.4	+15.0	-33.7	+37.1	+23.6	-15.4	0.0	OpenMHC team
4	DLinear	Neural	8.1	-11.5	-11.5	+27.2	-4.5	-15.4	+57.0	-58.6	0.0	OpenMHC team
5	LOCF (baseline)	Statistical	8.3	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
6	Temporal mode	Statistical	10.2	-12.2	-12.2	+45.7	-4.6	-20.5	-82.0	-19.2	0.0	OpenMHC team
7	Temporal mean	Statistical	10.6	-41.7	-41.7	-35.6	-19.6	+55.0	-6.6	-113.9	0.0	OpenMHC team
8	Mode	Statistical	10.6	-31.5	-31.5	+45.7	-2.9	-365.9	-82.0	-17.1	0.0	OpenMHC team
9	TimesNet	Neural	11.0	-79.1	-79.1	+6.2	-19.1	-207.4	-18.9	-125.6	0.0	OpenMHC team
10	FEDformer	Neural	11.4	-64.1	-64.1	+26.6	-16.6	-206.4	-66.5	-86.2	0.0	OpenMHC team
11	Mean	Statistical	13.4	-133.1	-133.1	-41.9	-26.4	-365.9	-82.0	-175.1	0.0	OpenMHC team
Long-context imputation (≥ 7×1440 time steps)
1	LSM-2-Sparse (7-day)	Neural	3.2	+63.1	+63.1	+40.3	+35.8	+91.7	+95.3	+30.5	0.0	OpenMHC team
2	LSM-2 (7-day)	Neural	5.4	+44.4	+44.4	+18.0	+22.4	+84.0	+89.6	+3.8	0.0	OpenMHC team
3	Pers. temp. mean	Statistical	8.9	-10.8	-13.5	+6.0	-1.0	+55.4	+4.8	-54.1	0.0	OpenMHC team
4	DLinear (7-day)	Neural	9.6	-36.7	-36.7	+17.6	-3.4	-48.0	+20.5	-86.8	0.0	OpenMHC team
5	Pers. mode	Statistical	10.4	-29.4	-29.4	+45.7	+1.1	-368.0	-82.0	-14.4	0.0	OpenMHC team
6	Pers. mean	Statistical	13.2	-120.8	-120.8	+0.1	-9.0	-429.7	-165.7	-146.3	0.0	OpenMHC team

Male · 1,331 users evaluated

#	Method	Type ▾ Neural Statistical	R ↓	S ↑	S_worst ↑	Activity ↑	Physio. ↑	Sleep ↑	Workout ↑	Semantic ↑	Fallback ↓	Submitter
Single-day imputation
1	LSM-2 (daily)	Neural	3.9	+62.1	+59.4	+40.7	+31.5	+90.3	+95.0	+32.1	0.0	OpenMHC team
2	Linear	Statistical	7.0	+21.8	+20.0	+4.2	+9.5	+61.9	+57.1	+0.3	0.0	OpenMHC team
3	BRITS	Neural	7.7	+9.3	+0.4	+20.3	-26.8	+40.7	+28.4	-1.6	0.0	OpenMHC team
4	DLinear	Neural	8.2	-3.8	-11.5	+29.9	-4.8	-9.4	+58.8	-42.0	0.0	OpenMHC team
5	LOCF (baseline)	Statistical	8.4	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
6	Temporal mode	Statistical	9.9	-3.4	-12.2	+47.1	+0.7	-10.4	-65.4	-7.9	0.0	OpenMHC team
7	Temporal mean	Statistical	10.3	-27.5	-41.7	-30.0	-17.8	+62.8	-1.2	-86.6	0.0	OpenMHC team
8	Mode	Statistical	10.6	-25.2	-31.5	+47.1	+0.2	-392.2	-65.4	-8.9	0.0	OpenMHC team
9	TimesNet	Neural	10.9	-62.5	-79.1	+10.3	-18.1	-222.5	+5.6	-97.0	0.0	OpenMHC team
10	FEDformer	Neural	11.2	-50.3	-64.1	+29.8	-13.5	-220.6	-49.3	-61.6	0.0	OpenMHC team
11	Mean	Statistical	13.4	-116.5	-133.1	-35.5	-24.9	-392.2	-65.4	-143.3	0.0	OpenMHC team
Long-context imputation (≥ 7×1440 time steps)
1	LSM-2-Sparse (7-day)	Neural	3.4	+65.3	+63.1	+41.5	+34.3	+92.2	+95.8	+36.2	0.0	OpenMHC team
2	LSM-2 (7-day)	Neural	5.6	+47.7	+44.4	+16.4	+18.2	+86.1	+91.0	+10.0	0.0	OpenMHC team
3	Pers. temp. mean	Statistical	9.2	-6.1	-13.5	+0.0	-7.2	+60.5	+19.0	-47.0	0.0	OpenMHC team
4	DLinear (7-day)	Neural	9.4	-25.8	-36.7	+20.3	-2.1	-37.5	+23.5	-64.5	0.0	OpenMHC team
5	Pers. mode	Statistical	10.6	-24.4	-29.4	+47.1	+2.2	-394.5	-65.4	-7.9	0.0	OpenMHC team
6	Pers. mean	Statistical	13.4	-112.2	-120.8	-5.0	-13.2	-445.7	-131.8	-128.7	0.0	OpenMHC team

Metric legend — scores are computed live vs the LOCF (last-observation-carried-forward) baseline.
R (Average Rank) — mean cross-method rank across all masking-scenario × channel tasks; 1 = best (lower is better).
S (Skill Score) — overall % reduction in reconstruction error vs LOCF (paired per-user geometric mean across tasks); higher is better.
S_worst (Worst-subgroup skill) — the method's lowest skill score across the four demographic cohorts (age <40 / 40+, female / male): its performance for the worst-served subgroup; higher is better.
Activity / Physio. / Sleep / Workout — per-category skill on that sensor group's channels (activity = steps, distance, flights; physiology = heart rate, active energy; sleep = asleep / in-bed; workout = the 10 workout-type channels).
Semantic — skill on the three structured-gap masking scenarios (sleep gap, workout gap, intensity failure).
Fallback — % of imputed values substituted by the LOCF baseline when the method produced no valid output (lower is better).
Source: MyHeartCounts/OpenMHC-leaderboard-data.
Cohort — the selector above recomputes R, S, and the per-category skills within one demographic cohort (a near-median age split at 40, or sex); the count above each table is the cohort's evaluated users. S_worst is the minimum cohort skill across the four cohorts, so it reads the same in every view. Cohorts are smaller than the full population, so scores are noisier; tasks without enough users (or degenerate within the cohort) drop from the aggregates.

Track 2b · Forecasting

All participants · 827 users evaluated

#	Method	Type ▾ Deep Learning Foundation Model Statistical	R ↓	S ↑	S_worst ↑	Activity ↑	Physio. ↑	Sleep ↑	Workout ↑	Submitter
1	Chronos-2 (fine-tuned)	Foundation Model	3.6	+37.6	+34.9	+30.7	+26.9	+63.9	+17.0	OpenMHC team
2	Chronos-2 (zero-shot)	Foundation Model	4.2	+36.4	+34.1	+30.5	+26.5	+62.3	+14.8	OpenMHC team
3	SegRNN	Deep Learning	4.4	+34.6	+33.4	+25.4	+20.8	+68.2	+2.5	OpenMHC team
4	DLinear	Deep Learning	4.6	+35.9	+33.8	+25.0	+16.5	+71.5	+5.5	OpenMHC team
5	Toto (fine-tuned, ctx4096)	Foundation Model	4.7	+30.9	+28.7	+29.5	+26.1	+46.1	+18.8	OpenMHC team
6	Toto (zero-shot, ctx4096)	Foundation Model	5.5	+26.8	+24.6	+29.2	+6.8	+50.5	+11.9	OpenMHC team
7	MixLinear	Deep Learning	5.7	+29.2	+27.7	+23.4	+13.4	+64.6	-7.2	OpenMHC team
8	AutoETS	Statistical	7.1	+14.3	+11.7	+0.6	-26.8	+37.6	+31.4	OpenMHC team
9	AutoARIMA	Statistical	7.6	+5.9	+4.1	-1.8	-9.0	+7.0	+24.0	OpenMHC team
10	Seasonal Naive (baseline)	Statistical	7.7	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team

Age < 40 · 358 users evaluated

#	Method	Type ▾ Deep Learning Foundation Model Statistical	R ↓	S ↑	S_worst ↑	Activity ↑	Physio. ↑	Sleep ↑	Workout ↑	Submitter
1	Chronos-2 (fine-tuned)	Foundation Model	3.7	+34.9	+34.9	+29.6	+25.7	+63.6	+5.8	OpenMHC team
2	SegRNN	Deep Learning	4.2	+34.4	+33.4	+25.2	+18.4	+68.2	+4.7	OpenMHC team
3	Chronos-2 (zero-shot)	Foundation Model	4.3	+34.1	+34.1	+29.3	+25.3	+61.9	+6.4	OpenMHC team
4	Toto (fine-tuned, ctx4096)	Foundation Model	4.6	+31.4	+28.7	+28.6	+25.5	+47.5	+20.5	OpenMHC team
5	DLinear	Deep Learning	4.7	+33.8	+33.8	+24.6	+13.1	+70.6	+0.3	OpenMHC team
6	Toto (zero-shot, ctx4096)	Foundation Model	5.6	+24.6	+24.6	+28.0	+4.7	+49.1	+7.6	OpenMHC team
7	MixLinear	Deep Learning	5.6	+27.7	+27.7	+23.1	+9.8	+64.7	-11.4	OpenMHC team
8	AutoETS	Statistical	7.1	+11.7	+11.7	-1.7	-25.6	+36.3	+25.3	OpenMHC team
9	Seasonal Naive (baseline)	Statistical	7.5	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team
10	AutoARIMA	Statistical	7.6	+4.1	+4.1	-1.7	-10.3	+5.0	+20.6	OpenMHC team

Age 40+ · 460 users evaluated

#	Method	Type ▾ Deep Learning Foundation Model Statistical	R ↓	S ↑	S_worst ↑	Activity ↑	Physio. ↑	Sleep ↑	Workout ↑	Submitter
1	Chronos-2 (fine-tuned)	Foundation Model	3.4	+40.3	+34.9	+31.5	+27.7	+64.2	+28.3	OpenMHC team
2	Chronos-2 (zero-shot)	Foundation Model	4.1	+38.7	+34.1	+31.3	+27.3	+62.8	+23.8	OpenMHC team
3	SegRNN	Deep Learning	4.5	+34.3	+33.4	+25.6	+22.4	+68.2	-1.3	OpenMHC team
4	DLinear	Deep Learning	4.5	+38.0	+33.8	+25.3	+18.8	+72.2	+12.7	OpenMHC team
5	Toto (fine-tuned, ctx4096)	Foundation Model	4.8	+30.6	+28.7	+30.1	+26.5	+45.5	+17.3	OpenMHC team
6	Toto (zero-shot, ctx4096)	Foundation Model	5.4	+28.1	+24.6	+30.1	+8.2	+51.6	+14.2	OpenMHC team
7	MixLinear	Deep Learning	5.7	+31.3	+27.7	+23.7	+15.6	+64.6	+2.0	OpenMHC team
8	AutoETS	Statistical	7.1	+16.6	+11.7	+2.4	-27.7	+38.4	+36.9	OpenMHC team
9	AutoARIMA	Statistical	7.7	+7.8	+4.1	-2.0	-8.2	+10.0	+27.1	OpenMHC team
10	Seasonal Naive (baseline)	Statistical	7.9	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team

Female · 195 users evaluated

#	Method	Type ▾ Deep Learning Foundation Model Statistical	R ↓	S ↑	S_worst ↑	Activity ↑	Physio. ↑	Sleep ↑	Workout ↑	Submitter
1	Chronos-2 (fine-tuned)	Foundation Model	3.7	+37.1	+34.9	+32.5	+26.6	+64.2	+11.7	OpenMHC team
2	SegRNN	Deep Learning	4.1	+37.8	+33.4	+26.2	+23.4	+70.2	+11.2	OpenMHC team
3	Chronos-2 (zero-shot)	Foundation Model	4.3	+36.2	+34.1	+32.3	+26.2	+62.7	+11.0	OpenMHC team
4	DLinear	Deep Learning	4.6	+36.8	+33.8	+25.6	+19.9	+72.9	+1.5	OpenMHC team
5	Toto (fine-tuned, ctx4096)	Foundation Model	4.7	+28.7	+28.7	+31.6	+26.6	+46.1	+4.6	OpenMHC team
6	Toto (zero-shot, ctx4096)	Foundation Model	5.4	+27.9	+24.6	+31.4	+7.0	+51.4	+13.0	OpenMHC team
7	MixLinear	Deep Learning	5.5	+31.7	+27.7	+24.1	+16.7	+64.6	+2.6	OpenMHC team
8	AutoETS	Statistical	7.1	+13.3	+11.7	+2.8	-23.0	+38.5	+23.2	OpenMHC team
9	AutoARIMA	Statistical	7.7	+4.5	+4.1	-0.5	-8.8	+11.8	+13.9	OpenMHC team
10	Seasonal Naive (baseline)	Statistical	7.8	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team

Male · 569 users evaluated

#	Method	Type ▾ Deep Learning Foundation Model Statistical	R ↓	S ↑	S_worst ↑	Activity ↑	Physio. ↑	Sleep ↑	Workout ↑	Submitter
1	Chronos-2 (fine-tuned)	Foundation Model	3.5	+37.3	+34.9	+29.9	+26.9	+63.7	+17.1	OpenMHC team
2	Chronos-2 (zero-shot)	Foundation Model	4.2	+36.2	+34.1	+29.8	+26.5	+62.1	+15.4	OpenMHC team
3	SegRNN	Deep Learning	4.4	+33.4	+33.4	+25.3	+19.5	+67.7	-1.3	OpenMHC team
4	DLinear	Deep Learning	4.6	+35.7	+33.8	+25.0	+14.9	+71.3	+6.3	OpenMHC team
5	Toto (fine-tuned, ctx4096)	Foundation Model	4.7	+31.6	+28.7	+28.8	+25.8	+46.1	+23.0	OpenMHC team
6	Toto (zero-shot, ctx4096)	Foundation Model	5.5	+26.2	+24.6	+28.3	+6.3	+49.9	+11.8	OpenMHC team
7	MixLinear	Deep Learning	5.7	+28.1	+27.7	+23.2	+11.7	+64.5	-11.0	OpenMHC team
8	AutoETS	Statistical	7.1	+14.0	+11.7	-0.0	-27.3	+36.4	+32.5	OpenMHC team
9	AutoARIMA	Statistical	7.6	+5.9	+4.1	-2.2	-9.2	+5.6	+25.8	OpenMHC team
10	Seasonal Naive (baseline)	Statistical	7.7	0.0	0.0	0.0	0.0	0.0	0.0	OpenMHC team

Metric legend — scores are computed live vs the Seasonal Naive baseline (24-hour-ahead forecasting; MAE on continuous channels, AUROC on binary).
R (Average Rank) — mean cross-method rank across channel tasks; 1 = best (lower is better).
S (Skill Score) — overall category-balanced % reduction in forecast error vs Seasonal Naive (paired per-user geometric mean); higher is better.
S_worst (Worst-subgroup skill) — the method's lowest skill score across the four demographic cohorts (age <40 / 40+, female / male): its performance for the worst-served subgroup; higher is better.
Activity / Physio. / Sleep / Workout — per-category skill on that sensor group's channels (activity = steps, distance, flights; physiology = heart rate, active energy; sleep = asleep / in-bed; workout = the 10 workout-type channels).
Fallback — % of forecasts substituted by the Seasonal Naive baseline when the model produced no valid output (lower is better).
Source: MyHeartCounts/OpenMHC-leaderboard-data.
Cohort — the selector above recomputes R, S, and the per-category skills within one demographic cohort (a near-median age split at 40, or sex); the count above each table is the cohort's evaluated users. S_worst is the minimum cohort skill across the four cohorts, so it reads the same in every view. Cohorts are smaller than the full population, so scores are noisier; tasks without enough users (or degenerate within the cohort) drop from the aggregates.

Submit your model

Add a method by opening a pull request on the OpenMHC leaderboard dataset that adds your per-user evaluation substrate (<track>/<method>.parquet) plus a small <method>.meta.json sidecar. Produce the substrate by running the OpenMHC eval with output_dir=…; the maintainers recompute the skill, rank, and worst-subgroup scores from it. See the step-by-step submission guide for the exact file schema.