Benchmark

BIG-Bench Extra Hard

What is BIG-Bench Extra Hard?

Successor to BBH after saturation; per-task variance is high. Benchmark Radar tracks 1 reported scores for BIG-Bench Extra Hard, cited by 1 source documents…

Released
2025-02-26
Evaluates
reasoning
Openness
unknown
Importer
Model reports
Reported scores
1

Source provenance

BIG-Bench Extra Hard paper

BIG-Bench Extra Hard results and reported scores

Model reports

ModelOrganizationReported valueReportedEvidence
Gemma 4 (31B)Google74.42026-03-11evidence

Scores are partitioned by the source that reported them and are never merged into a single cross-source ranking, because the sources measure different things and say so.

Which sources cite BIG-Bench Extra Hard?

Related reasoning benchmarks