DECIDE

Hosted evaluation

Evaluate your model.

Submit a model API. Our worker runs the private benchmark and returns aggregate accuracy by task, decision family and clip duration.

Dataset files and reference labels are not provided. Your model API operator receives the clips or transcripts and context sent for inference.

1. Submit an endpoint

Start directly with your HTTPS model endpoint. No invitation, email or manual approval is needed. Full evaluations automatically check your API on a public example before using private inputs. Read the API contract. Automated daily and concurrency limits apply.

Model API keys are not saved in this browser. Only the private result receipt is saved on this device. Model providers may charge for inference; the full run sends one request per question, with bounded retries. A full run also sends one public example to check API compatibility.

2. Follow your evaluation

Jobs continue after you close this page. Keep the private receipt to check results on another device. Anyone with the receipt can view or cancel that evaluation.

No evaluation saved on this device yet.