General Knowledge

A 70-item BIG-bench multiple-choice set of child-level and oddball English facts, scored as multiple_choice_grade.

Also known as: BIG-bench general_knowledge

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.
Categoryknowledge
Subcategoryshort English common-sense and trivia multiple choice (BIG-bench)
Page statusunknown
Metricmultiple_choice_grade
Directionhigher_is_better
Unit%
Dataset size70
Dataset licenceApache-2.0
PublisherGoogle (BIG-bench collaboration)

What it measures

general_knowledge asks short English questions that a child could often answer, plus a few absurd or slightly harder facts. The author, Evgenii Zheltonozhskii, groups items as "obvious" (how many legs a horse has), "absurd" (how many tails a human has), and ordinary trivia that is likely somewhere in pretraining. The point is coherent factual answering, not reading a passage. It is not [MMLU](mmlu.md) and not MMLU Global Facts.

Task format

Multiple choice, preferred metric multiple_choice_grade. append_choices_to_input is true. Option count varies: 4 to 13 choices (mean 7.57; 16 items have seven). Canary GUID embedded. Dummy-model header: 70 multiple-choice and 0 free-text queries. Redesigned from exact string match after reviewer discussion.

Models reporting this benchmark

No model card in ModelSpec reports this benchmark yet.

Data

This page as JSON · Edit on GitHub