ARGIRA IX extends the ARGIRA VIII framework on cross-modal invariance analysis by decomposing the aggregate acoustic target tempo_bpm into three interpretable perceptual components: chromatic (tempo_color), structural (tempo_fractal), and spatial (tempo_space) channels. The dataset investigates whether visual-to-acoustic associations persist under explicit operator-induced reparameterization of the mapping function across three configurations (M1–M3), which include baseline, functional transformation, and architectural reorganization of feature-to-target assignments. Results indicate that associations previously identified as invariant under aggregated targets do not persist when evaluated at the level of decomposed acoustic channels. Instead, predictive signal structure exhibits systematic redistribution across operator-defined channels, suggesting strong dependence on representational architecture rather than intrinsic invariance of the underlying corpus. The release includes full analysis code, operator definitions, classification taxonomy, and visualization artifacts supporting reproducibility of the invariance testing pipeline.