METAFOR exploratory landscape analysis features for continuous optimization benchmarks (dimensions 10 and 50)

This dataset provides exploratory landscape analysis (ELA) features for continuous-optimization benchmark instances evaluated through METAFOR in dimensions 10 and 50. The instance registry contains 456 problem-dimension records. Four records with unbounded search domains were intentionally excluded from sampling, leaving 452 usable records. Each usable record was sampled in 100 independent repetitions, producing 45,200 successful per-repeat feature observations and 452 aggregate records. The dataset contains portable sampling manifests, raw and standardized per-repeat ELA features, per-instance aggregate means and standard deviations, scaler parameters, extraction metadata, checksums, and validation reports. A total of 107 ELA feature columns were extracted using flacco; 99 contain at least one finite value. Raw sampled decision vectors, objective-value files, intermediate feature chunks, and visualization outputs are intentionally excluded from this compact core dataset. Sampling commands, seeds, point counts, statuses, elapsed times, software provenance, and excluded-data details are recorded in the included manifests.