Internal A/B evaluation surveys for model responses.

Build comparison surveys, collect calibrated win-rate data with Wilson 95% confidence intervals, and run AI follow-up probes — all gated to your team’s work email.

• 7 question types

  A/B compare, rating, single & multi choice, short & long text, email. CSV bulk import for response pairs.

• Calibrated results

  Per-dimension win-rate with Wilson 95% CIs, completion histogram, raw & structured CSV exports.

• AI follow-ups

  Optional gpt-4o-mini probe per question, served via Supabase Edge Function with a per-survey daily call cap.