Ah, guess I should have looked at the paper. I foolishly assumed that if they had an actual big-O improvement over Strassen, they’d mention that important fact in the abstract and blog post.
Any improvement over Strassen on 4x4 matrices represents an asymptotic improvement over Strassen.
ETA: this claim is still right, but they only asymptotically improve Strassen over F2, not for normal matrix multiplies. So your intuition about what they would definitely mention in the abstract / blog post may well be right.
Ah, guess I should have looked at the paper. I foolishly assumed that if they had an actual big-O improvement over Strassen, they’d mention that important fact in the abstract and blog post.
Any improvement over Strassen on 4x4 matrices represents an asymptotic improvement over Strassen.
ETA: this claim is still right, but they only asymptotically improve Strassen over F2, not for normal matrix multiplies. So your intuition about what they would definitely mention in the abstract / blog post may well be right.
Indeed. Unfortunately, I didn’t catch that when skimming.