TenPeers · A peer-ranked journal for the AI age

Sitting on a paper?

Send us the paper you still believe in. Before you fold it into a chair.

Ten peers rank your paper. Decision in 4 weeks.

Use your current PDF. No reformatting.

First cohort. Submit by 31 December 2026, 23:59 UTC · Decisions by 31 January 2027 · Dates and fees

A lecture-hall chair folded from the pages of a two-column manuscript stands on an oxblood floor. A red pencil tick marks the backrest.
Object study, paper chair

For ICLR authors

Three reviews.
Feeling lucky?

Authors
We proved the Riemann hypothesis.
Reviewer 2
Weak reject. No comparison to baselines (Riemann, 1859).

Adapted from Ravid, Cool AI Guy at Meta

A fresh read for your ICLR paper
A three-reel slot machine folded from a research manuscript. Its paper reels read 1, 2 and 3.

Why TenPeers exists

The papers still in the drawer.

“In one conference cycle, our team submitted seven papers. We had seven more from earlier rounds that we still wanted people to read. Every time we picked one up, we had to choose a venue, reformat it and start the wait again. That’s why we built TenPeers.”

Mikhail Arbuzov, founder

A shallow drawer folded from printed ivory manuscript pages stands halfway open on clay. Six manuscripts sit inside. A seventh projects over the lip, with a pencilled correction changing "mean" to "median."

One monthly cohort

What happens after you submit?

  1. Approve your summary.

    We draft it from your PDF. You edit and approve what peers will read.

  2. Rank ten papers.

    Your team ranks ten other summaries twice, once for research impact and once for talks you'd attend.

  3. Read the finalists.

    About 30% advance on impact. One author from each finalist team reviews 3–5 other papers in full per finalist submission.

  4. Get your decision.

    Your letter explains the outcome. For finalists, it draws on both impact rank and full reviews.

  5. Publish.

    If accepted, your paper goes online in TenPeers Select or TenPeers Research, at a fraction of the usual cost.

What gets cited?
What fills a room?

Which papers will researchers build on and cite? Which talks would you go to? Rank the same ten summaries twice.

  1. Expected impact

    Which papers will other researchers build on and cite?

    Two-year horizon. Decides who goes to full review.

  2. Talk interest

    Which talks would you go to in person?

    Goes to event organizers. Never to a publication decision.

One peer, two separate judgments, kept apart until both are committed.

Illustrative

Follow one paper

One paper can find two audiences.

Take When should a small model ask for help? Its authors study when a cheap model should hand a difficult case to a larger one. Consider what two different readers might want from that result.

In the journal

Placed 8th of 100 for expected impact

A reader designing an inference system wants to cite the experiment and compare its threshold with their own. Eighth for impact takes the manuscript through to full peer review.

In the room

Placed 2nd of 100 for talk interest

Another reader wants to ask what happens when the traffic changes overnight. Second for talk interest gives a local organizer a reason to put that question on a workshop program.

In an illustrative cohort of 100, the paper places 8th for expected impact and 2nd for talk interest.

Two researchers made from printed ivory paper lean in to point at different parts of the same attention map.

The TenPeers Society

On one paper you're the author.
On the next, you're a peer.

The people who submit the papers are the people who choose them. See where one reader placed one paper, twice.

First cohort · submit by 31 December 2026, 23:59 UTC

Have a paper in mind?

Use the PDF you have. You'll see your fee and decision date before you confirm.