Benchmark Results

This section contains three categories of benchmark results:

• 3rd Party Solvers — max-div against other freely available subset-selection tools: Python heuristics, exact MIP/CP solvers, and the MDPLIB best-known values from the literature.
• Solver Presets — max-div's own initialization/optimization strategies and presets measured against each other on the built-in test problems.
• Internal Primitives — micro-benchmarks of internal building blocks (diversity-metric kernels, random samplers, probability selectivity).

The Solver-Preset and Internal-Primitive benchmarks are triggerable via the CLI:

max-div benchmark <category> <sub-command>

All benchmarks are run on a 16" MacBook Pro with M3-class CPU.