What it is
Benchmark of 4,084 ten-choice health QA items grounded in real wearable histories, biomarkers, and demographics from 200 users. 4,084 ten-option QA examples from 200 real users, each with up to ~500 days of daily wearable data plus biomarkers and demographics, released in 5 text/structured serializations and 16 question types.
Key results
- 4,084 ten-option questions
- 200 real users with up to ~500 daily measurements each
- 5 representations; 16 question types; CC BY-NC 4.0
Useful for benchmarking QA and reasoning models on real-world wearable health time series.