New project

Upload two evidence-replay CSVs: a reference window and a comparison window. Each row needs a score column (0–1). An optional label column (0/1, blank if unavailable) enables discrimination and calibration. Optional attr_* columns (per-feature attribution magnitude) enable explanation-stability comparison.

  1. 01Declare execution contract
  2. 02Validate two windows
  3. 03Compute evidence
Execution capability

Leave the manifest empty for Evidence replay, or attach a safe, data-only logistic manifest to rerun a supported model from raw-feature CSVs. Executable Python, pickle and joblib are never accepted.

Evidence replay · supplied scores/attributions · model is not rerun

The baseline/earlier window this project evaluates against.

Drag and drop a .csv file here, or click to browse

score (required), label (optional), attr_* columns (optional) · max 5MB

The later/operational window compared against the reference.

Drag and drop a .csv file here, or click to browse

score (required), label (optional), attr_* columns (optional) · max 5MB

Select a valid reference and comparison CSV above to enable this. Nothing is uploaded until you press this button.