START HERE

This archive reproduces the fixed-size real–synthetic mixture benchmark reported in the submitted manuscript. Every classifier-training scenario contains exactly N rows (RR, RS25, RS50, RS75, SR). The historical RR/SR/RS 2N arm was not used in the primary fixed-size benchmark. The distinct RR-N/RB-2N/RSADD-2N incremental control is documented in Supplementary File 4.

Read, in order:
1. README.md — scientific scope and scenarios
2. README_REPRODUCIBILITY.md — commands and expected outputs
3. EXPERIMENT_PROTOCOL.md — locked protocol and interpretation boundaries
4. LICENSE_AND_ATTRIBUTION.md — dataset licences and citations
5. CHECKSUMS_SHA256.txt — archive integrity
