t-SNE embeddings for the original targets (a–c) and the reduced targets with RUWE < 1.4 (d–f). Panels (b) and (e) show the t-SNE visualisations annotated with known classes from various sources: Gaia classifications (legends in the bottom left), SIMBAD (white dwarfs, labelled as WD_SB), Paper I (Hsd_C0, Hsd_C1, CV_C2), and CVs from the literature (CV_Lit). Panel (c) displays cluster labels derived from a Gaussian mixture model, where clusters are labelled according to known object types rather than numerical identifiers. Panel (f) shows the same cluster labels in panel c for the reduced dataset. The SOS and ML annotations in the legends refer to objects classified from the Gaia SOS and ML pipelines, respectively (see also Fig. 3).