๐พ CMT Semantic Alignment Lab
Can CMT phase-geometry decode what dogs are saying?
This app runs a rigorous scientific experiment using the CMT/SRL/SEFA framework to test whether the information geometry of dog vocalizations overlaps with semantically equivalent human emotions โ and whether that overlap is statistically distinguishable from chance.
Every result is labelled with its epistemic status: [SETTLED], [CONFIRMED], [HYPOTHESIS], or [SPECULATION].
Dataset Summary (REAL DATA โ )
| Human | Dog | |
|---|---|---|
| Samples | 1440 | 776 |
| Labels | angry, calm, disgust, fearful, happy, neutral, sad, surprised | bark, growl, howl, whine |
| Features | 10 CMT features | same |
| Lenses | 4 (Airy, Bessel, Gamma, Zeta) | same |
| NaNs | 0 | โ |
CMT Alpha Airy (best-performing lens per monodoc):
- Human mean ฮฑ:
0.6553ยฑ0.1760 - Dog mean ฮฑ:
0.5383ยฑ0.2228
Note: ฮฑ values for Gamma and Zeta lenses are systematically low across both species, consistent with the V2 monodoc (Zeta low-ฮฑ known limitation, Gamma near-constant for these input ranges). Airy and Bessel are the informative lenses for this dataset.