Blog Post··6 min read
Formatting & Output Validation Benchmarks: IFEval Is Saturated, IFBench Isn't
IFEval's top-5 span fewer than 2 percentage points — frontier models have converged on its constraint set. IFBench exposes a 29pp gap between Grok and Claude on out-of-distribution constraints. And SOB shows that JSON schema compliance is not the same as correct field values.