¡NOTE! *New Update* (04/09/2025) Added ability to test style models and fixed gui path variables. Please see Civitai/GitHub pages for more details. (>'.')>[<3]

_____________

A "General Accuracy Score" (GAS) has been developed to quantify a model's ability to recreate the likeness of a trained subject, as compared to another one with similar hyperparameters (training or image generation). Statistical weighing methods are leveraged to characterize different important aspects of a given model's data distribution.

_____________

A "Mean General Accuracy Score" (MGAS) expands upon the previous, taking the results from the various weight-based-GAS characterizations and combining them through the Mean-GAS. More than two models can be compared through a Multi-Model Round-Robin tournament.

_____________

Please check the article out and let me know what you think! ദ്ദി(ᵔᗜᵔ)