List your company

Question Answering on BIG-Bench Hard (BBH) (Specific Metrics)

55.96BBH MCQ AccuracyMCQ-Matched
54.233654.681855.1355.5782Jun 10, 2026

Evaluation results

MethodLinks
MCQ-MatchedCondition=MCQ-Matched2026.0655.9662.6322.3
Full-Format SFTCondition=Full-Format2026.0655.2362.2824.6
Random-FormatMixCondition=Random-FormatMix2026.0655.0660.6823.5
Targeted-FormatMixCondition=Targeted-FormatMix2026.0654.9360.5424.5
GLM4-9BCondition=Baseline2026.0654.362.1423.4