New benchmark measures LLM consistency for structured outputs

A developer has released a benchmark specifically designed to test how reliably LLMs produce deterministic, structured outputs—a practical concern for anyone building production systems that depend on consistent formatting. This addresses a real gap: existing benchmarks focus on reasoning or accuracy, not reproducibility, which matters when you're integrating LLM outputs directly into code.