BenchScope: How Many Independent Signals Does Your Benchmark Provide?
This work addresses the issue of inefficient evaluation for AI researchers and benchmark maintainers by providing a method to identify redundant measurements, though it is incremental as it builds on existing diagnostic concepts.
The authors tackled the problem of redundant scores in AI benchmarks by introducing Effective Dimensionality (ED) as a diagnostic tool, revealing that many benchmarks, such as the Open LLM Leaderboard with six scores behaving like roughly two effective axes, show substantial redundancy across 22 benchmarks and over 8,400 model evaluations.
AI evaluation suites often report many scores without checking whether those scores carry independent information. We introduce Effective Dimensionality (ED), the participation ratio of a centered benchmark-score spectrum, as a fast, population-conditional upper-bound diagnostic of measurement breadth. Applied at per-instance granularity to 22 benchmarks across 8 domains and more than 8,400 model evaluations, ED reveals substantial redundancy: the six-score Open LLM Leaderboard behaves like roughly two effective measurement axes (ED = 1.7), BBH and MMLU-Pro are near-interchangeable (rho = 0.96, stable across seven subpopulations), and measurement breadth varies more than 20x across current benchmarks. We show that relative ED rankings are stable under matched-dimension controls and that ED can flag redundant suite components, monitor performance-conditional compression, and guide benchmark maintenance. Because binary spectra overestimate absolute latent dimensionality, we interpret ED as a screening statistic rather than a literal factor count and complement it with null, reliability, and saturation analyses. We provide a 22-benchmark reference atlas and a four-step diagnostic workflow that benchmark maintainers can run with a score matrix and a few lines of code.