I had hoped that would be the case, but I tried multi-lingual clustering, and the clusters were neatly assigned per language, even though the meanings were incr

I had hoped that would be the case, but I tried multi-lingual clustering, and the clusters were neatly assigned per language, even though the meanings were incredibly close (e.g. short phrases literally meaning "cheap" or "unhealthy"). BTW, that was using ada-002.