When measured human motion reaches a model, mismatched coordinates, units, or timing can produce an answer that runs but means the wrong thing. This program studies how to protect that boundary with executable checks rather than human attention alone.
The question
Which input contracts, dataset boundaries, and failure conditions can stop coordinate mix-ups, target leakage, and false external validation before a model runs?
Current public state
The interface contract rejects seven deliberately broken variants. The dataset gate splits Carter into 30 development, 10 validation, and 10 test participants; reserves OpenCap laboratory data for external validation; and reserves Knee Grand Challenge for internal-load stress testing. B3D is a source-adapter format, the internal representation is NumPy plus explicit semantics, and Nimble is not mandatory.
Evidence boundary: No model has been fitted and no holdout has been read.
Read the current Research Note (EP-0005) →