> Multiverse turned to a mathematically complex approach borrowed from quantum physics that uses networks of high-dimensional grids to represent and manipulate large data sets. Using these so-called tensor networks shrinks the size of the model significantly and allows a complex AI system to be expressed more efficiently.

> The method gives researchers a “map” of all the correlations in the model, allowing them to identify and remove specific bits of information with precision. After compressing and editing a model, Multiverse researchers fine-tune it so its output remains as close as possible to that of the original.

This seems to be the substance but I didn’t spot a link to a paper. Is there a technical explanation someplace?

I think this is the paper? Skimming through it now: https://arxiv.org/pdf/2401.14109

edit: it's really light on details. They have some graphs on reduction and a few (old) benchmarks where supposedly they don't lose much accuracy, but with such old models being listed, it's hard to know. More of a "promo pamphlet" than a paper tbh.

Is this really the right paper. It's from May 2024.

Why does this site attempt to force me to download some .htm file?

bad mime rules on the server