Generalizing Transformation-Based Interpretation Beyond Pitch Transposition

Determine how transformations other than pitch transposition can interpret forms of music beyond the predominantly Western concepts induced by the current twelve-tone transposition setting.

Background

The framework relies on one-semitone pitch transposition as its principal inductive bias and is evaluated largely on data representing Western tonal music. The authors note that the emergence of twelve-feature rings may be tied to the twelve-tone equal-temperament structure and the distribution of the training data. An unresolved direction is therefore to identify suitable alternative transformations and determine what musical structures they reveal in other musical traditions or forms.

References

Additionally, although 12-sized rings emerge from unsupervised SAE alignment, it is potentially related to the pitch transposition step size being $1/12$ of an octave. That, combined with the distribution of data we use, yields mostly Western music concepts. How other transformations may interpret other forms of music is yet to be studied.

— From Isolated Feature to Orbits: Discovering Music Concepts via Multi-SAE Alignment  (2610.01864 - Lin et al., 1 Oct 2026) in Section 6, Limitations