A multi-modal infant-based metric for choosing the best supernova

Abstract

We present a comparative study of 22 core-collapse supernovae (SNe), selected to explore a novel, multidimensional ranking scheme aimed at identifying the best supernova. Each SN is evaluated based on three principal criteria: (1) inferred explosion energy derived from light curve modeling and spectroscopic indicators; (2) an aesthetic score assigned to the SN host galaxy following transformation into a human face using a generative visual model (Midjourney v5); and (3) final ranking by Claud.IA, a 6-month-old infant trained to select the best SN via repeated exposure to curated SN images. We define and normalize all criteria to ensure statistical consistency across the sample, with particular attention paid to the biases inherent in infant-based classification models. The top five SNe exhibit both high explosion energies (E > 1e51 erg) and extremely cool host galaxies (post transformation), with Claud.IA showing strong preferences toward galaxies exhibiting symmetric facial morphology and prominent spiral arms. Final application of Claud.IA identified the best supernova in our sample as SN 2022joj. Our study demonstrates the feasibility of incorporating human-machine hybrid aesthetic judgment and early developmental cognition into astrophysical classification, and raises intriguing questions about the nature of bestness in cosmic explosions. Additional follow-up is encouraged.