FAQ

No, and there is no mode where it does. The panel produces an advisory read per submission. A person sets the Jury Score per dimension, and the leaderboard is built from those human scores and your criterion weights. The AI Total Score sits next to the human one for reference and never ranks anyone.

It reads the deck in full, scores it against your rubric with a panel of independent AI judges, ties every finding to the passage it came from, and hands your team a report plus a ranked board. It never issues the decision.

A batch is evaluated in a single run rather than deck by deck, so a cohort that would take a team weeks of reading comes back as a comparable set. The pace that matters afterwards is your review meeting, not processing time.

No. You can add entrants by hand or open a public submission page through the Entry Hub, and existing intake tools keep working. Applicants do not change what they send.

Yes. Criteria and weights are yours, and they apply at the leaderboard rather than inside each judge's reading — so the same evidence can be re-ranked without re-running the batch. They lock when scoring starts, so the whole field is ranked on one standard.

Tracks, criteria and weights are configured per project, and each track scores against its own rubric. The leaderboard respects the weighting of its track.

Never trained on — contractual. Submissions are processed only for your program's evaluation, inside a workspace scoped to your project. The sub-processor and model-provider list is published for IT review, and every report and score belongs to your program.

State it before they ask. The honest framing is stronger: every submission received a full read under identical rules — which no volunteer process can claim — and humans made every ranking decision. What is genuinely hard to defend is entry #300 getting a tired judge at 11pm.

Nothing here cuts a judge. Judge count is a sponsor perk and a program KPI, so what changes is the ask, not the headcount. Remote screeners keep their touchpoint as a confirm-and-comment pass instead of a reading weekend, and they say yes again next year.

Every submission has a record: scores per dimension, the evidence and quotes behind them, where the reviewers disagreed, and the human score that decided the placement. Whether you run open appeals stays your policy — the record is what turns that conversation into five minutes instead of an archaeology dig.

Run it in parallel. Your screeners work a round as usual while the panel reads the same field; you compare coverage, hours and where the pre-ranking agreed with your screeners. Nothing counts until you decide it does.

Not today, and we do not claim otherwise. The panel reads the submission materials you already collect — the deck, the project description, the team's notes. Repository and live-URL review is the next build on the roadmap.

No. Deck completeness flags missing or thin coverage; it does not validate a claim. Prompt-injection safety prevents deck text from controlling the evaluation; it does not prove the deck is truthful. Both boundaries are in What EvalLens does not do.

One submitted deck that receives one successfully generated AI evaluation report. A deck that cannot be processed does not count, and neither does disagreeing with the output. A resubmission after a report is a new evaluation.

There is no public sign-up — EvalLens runs a limited partner program, and access is set up for your team after a call. Ask on that call what a first-run offer looks like for your program.