Muse Spark reads a CSV of market data, flags duplicates, gaps, bad prices and unadjusted splits, and explains each one in plain English. A deterministic scorer then checks its answer.
Muse modelsA task passes only when the problem set is exactly right (F1 = 1.0) and the overall score is ≥ 0.8, with numeric fields within tolerance.
A seeded random walk produces clean prices. Known problems are planted at recorded rows, which becomes the answer key.
The CSV goes to a Muse model, which returns each problem it finds with the affected rows and an explanation.
The scorer counts caught, missed and falsely flagged problems against the answer key and returns one number.