How we score every paper

Every story on Just Out Tech carries a credibility score out of 10. It is printed at the top of the article and on every card, next to the reading level. This page explains how the score is built, so you can check our work.

The five parts

Peer review and venue (0 to 3). Three points for a paper published in a top venue (a leading journal or an A-ranked conference). Two for a paper accepted and in press. One for a preprint from a strong lab. Zero for a preprint we cannot place.

Labs and authors (0 to 2). Two points for institutions or labs with a track record in this exact field. One for credible authors. Zero when we cannot verify who wrote it.

Evidence type (0 to 3). Three points for a real-world or hardware result with proper baselines and no cherry-picking. Two for a controlled experiment or simulation. One for a survey or framework. Zero for opinion.

Open code and data (0 to 1). One point when the code or data is public and linked.

Community signal (0 to 1). One point for notable citations, attention from other researchers, or an independent replication.

What the score is not

It is not a judgment of the authors. A new preprint from a good lab will often score 6 or 7 simply because nobody has had time to check it. The score goes up when the paper is published or replicated, and we update it.

Our rules

Nothing is published from a summary. We open the paper, find the number, and note the page. Preprints are labelled “not yet peer reviewed”. We compare every press release with the paper it describes. We check the Retraction Watch database. Corrections are published, dated and linked, never silently edited.