Definition
Plain language
A formal withdrawal of a published paper because it turned out to be wrong or dishonest.
As stated in the literature
Editorial removal of an article from the scientific record for error, misconduct, or unreliable data; retraction status is the primary credibility signal TRACES tests for.
Also called: retracted, retractions
Why it matters: It is the clearest available signal that a result should not be built on, so a system that ignores it will keep spreading claims the field has already discarded.
For example, a journal can pull a paper years after publication once an investigation finds its images were duplicated.
Heard on the show
“They sorted completions by worst prediction error and found the retraction pattern there.”Episode 204 — The Length Estimate Hiding Inside a Word-by-Word Model