RadioML 2018.01A for open-set recognition (OSR) in RF ensembles is a gold-standard synthetic RF for modulation recognition, ideal for OSR experiments. Treat known classes (e.g., BPSK, QAM) closed-set taxonomy and hold out subsets (e.g., noise, unseen modulations) as "Unknown" for abstention testing. Reproducible, SNR-stratified, integrates seamlessly where we evaluate thresholding (max-prob, entropy, energy) with OSCR curves and AU-PR on synthetic benchmarks with held-out unknowns.

