Figure 3

Trends in WM detection accuracy across Experiments 1 and 2. Visualization of classification accuracy trends (Tables 3 and 4). $\mathrm{\Psi }\left({\tilde{\mathbf{X}}}_{GE}\right)$ (consensus embedding) performs significantly better than comparative strategies (original feature space, GE, MDS); using median as ML estimator (purple) may be marginally more consistent than using mean as ML estimator (orange). Ψ(F) (blue) and Ψ(X _MDS ) (red) perform similarly (corresponding trends directly superposed on one another).