Model Construction

Model Lab

Train the two list-level candidate models, compare their feature reliance, score completed jobs, and upload site groups to see how trained models behave across the domains you have already scanned and labeled.

Overview

Loading model overview...

Train Variants

Training uses human-reviewed candidate rows only. Leave the job filter blank to use all labeled candidates currently stored, or limit training to specific jobs by entering comma-separated job IDs.

Trained Models

Use Runtime JSON when another browser component needs a compact deployable inference bundle. The full saved artifact remains available through the API for research and debugging.

No trained models saved yet.

Score Existing Job

Use a trained model against a completed job. This scores every stored candidate on the job, re-ranks each page, and shows whether the top candidate looks auto-acceptable or still needs manual review.

Loading recent jobs...

Run a scored job to see item-level predictions.
No scored job loaded.

Site Group Probe

Paste or upload URLs or hostnames. The probe matches those sites against pages you have already scanned, scores them with the selected model, and groups the results by hostname so you can compare how the model behaves across site families.

No site-group probe has been run yet.