A Deterministic Benchmark for Physics-Impossibility Detection via Method-Level Falsifiability

This record archives Benchmark v2 (Frozen), a deterministic and reproducible benchmark designed to evaluate whether a computational decision system can classify structured physics scenarios as Impossible, Admissible, or Indeterminate using method-level falsifiability rather than probabilistic inference, statistical learning, or empirical fitting. The benchmark consists of a permanently frozen dataset, a deterministic decision gate, and a reproducible execution pipeline. All benchmark assets are cryptographically hashed and time-stamped. Identical inputs are required to produce bitwise-identical outputs. Mismatches are explicitly preserved and treated as first-class data. This benchmark does not claim validation of physical theories, discovery of new physics, or replacement of experimental science. It evaluates methodological coherence under a declared constraint envelope.