Authorship Verification (BIG-bench)

BIG-bench two-choice task: match a ~500-word Gutenberg passage to its author, in same-genre and genre-swapped pairings.

Also known as: Authorship Verification in a Swapping Scenario

unassessed

This page is a discovery lead. Nobody has yet assessed it against the catalogue contract, so it carries no disposition. Absence of evidence here is not evidence of staleness.
Categoryreasoning
Subcategorystyle-based same-author identification with genre swap
Page statusunknown
Metricmultiple_choice_grade
Directionhigher_is_better
Unit%
Dataset size880
Dataset licenceApache-2.0
PublisherGoogle (BIG-bench collaboration); task authors at HU Berlin, Uni Weimar, Uni Halle, Uni Leipzig

What it measures

The model sees one reference passage of about 500 words and two candidate passages of similar length. It must pick the candidate written by the same author as the reference. One subtask keeps genre aligned (adult vs children's literature). The swapped subtask pairs same-author texts across genre and same-genre texts across authors, so topic cues point the wrong way.

Task format

Two-option multiple choice. Preferred metric multiple_choice_grade. Dummy-model header counts 880 queries. Canary GUID embedded.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub