Pick a template. Click one emoji. Done. The form expands only when you flag something. Real eval lands in our dashboard - refresh it after you submit.
After you submit, open the live dashboard → and you'll see your eval landed.
Progressive disclosure isn't just a buzzword. It's the difference between a form people fill and a form people abandon.
Every choice in this form is research-backed. Three that matter most:
Anchoring only score 1 and score 5 gives equal human-judge alignment as anchoring all five - at lower cognitive load.
For LLM judges sampling N=5 times, mean aggregation beats median and majority across every model and decoding strategy tested.
For more than two raters with mixed scales and missing data, α is the right reliability metric. We compute it live on the dashboard.
Yes. Open. No signup, no card, no rate limit. Your submissions land in the shared demo store, visible on the dashboard. If you want private storage, see the install page for self-hosting.
One script tag. One element. Install the adapter SDK → for copy-paste integration, postMessage protocol, and live API reference.
Not on this page - you're the human rater. To trigger LLM-judge fills, POST a JSON payload to /api/save_eval.php with rater.type: "llm_judge". Same schema as the form.
Open the dashboard. Latest evals at top with rater, modality, score, safety verdict, and tags.
Yes. The template config is in form.html - a 100-line JS object. Add a new market, tune the headline, choose which dimensions show in Standard. Same schema underneath.
You just used the form. Drop the same form into your SaaS or AI product with one script tag.