{"id":"palomar/PALOMAR-2026-11-14-000006","url":"https://github.com/arhancanli/quantavane-fmm-9x11x11/tree/f3eb2873d0e27364a75846568294a566cc1fa2b7","json":"https://data.palomar-registry.org/entries/PALOMAR-2026-11-14-000006-v1.json","repo":"https://github.com/arhancanli/quantavane-fmm-9x11x11","tags":["cs.SC","cs.CC","msc:68Q25","msc:15A69","msc:68W30"],"score":0,"title":"Multiplying 9 × 11 by 11 × 11 matrices with 708 multiplications","trust":"high","domain":"cs.SC","source":"palomar","authors":["Arhan Canli"],"reviews":0,"subject":"cs.SC","summary":"A Lean 4 and Mathlib proof that the matrix multiplication tensor ⟨9, 11, 11⟩ has rank at most 708 over the rationals. Two compared theorems: 708 triples of rational matrices decompose the tensor (∑ᵢ uᵢ[a, b] vᵢ[c, d] wᵢ[e, f] = [b = c ∧ a = e ∧ d = f]), and the same triples form a bilinear algorithm A * B = ∑ᵢ (∑ uᵢ[a, b] A[a, b]) (∑ vᵢ[c, d] B[c, d]) wᵢ over every field of characteristic zero. The scheme, scaled to integers, is data; Lean's kernel checks it (decide +kernel, no native_decide) slice by slice with the coefficients packed into large integers, and a soundness proof for every…","category":"research","comments":0,"disputed":false,"division":"cs","mentions":0,"verified":true,"discussed":0,"published":"2026-11-14","reactions":{},"formalized":true}