Deterministic combinatorial Boolean matrix multiplication

The Boolean product records whether an entry in a row and the matching entry in a column are both 1 for at least one position. How fast can a deterministic combinatorial algorithm compute the entire product?

Why it matters and what counts as progress

Why it matters: Boolean products encode two-step reachability. Related reductions connect faster products with faster triangle detection and other graph problems.

What counts as progress: A faster algorithm needs an exact correctness proof and a worst-case bound in a stated model of computation. The same model is needed to compare running times.

How the problem developed

  1. 2024 · Graph decomposition [5]

    Abboud, Fischer, Kelley, Lovett and Meka Gave a deterministic combinatorial algorithm with a superpolylogarithmic saving over cubic time.

  2. 4 Nov 2026 · Faster deterministic algorithms [1][3]

    Cui, Liu and Woodruff with Gemini 4 Reached the square-root exponent in the saving for Boolean products and triangle detection.

AI contribution

Combinatorial Boolean Matrix Multiplication in n3/2Ω((log⁡n))n^{3}/2^{\Omega }(\sqrt{(\log n)}) Time

Piaoyang Cui, Yiming Liu, David P. Woodruff

Cui, Liu and Woodruff give deterministic combinatorial algorithms that multiply two n×nn\times n Boolean matrices or decide if an nn-vertex graph contains a triangle in O(n3/2Ω(log⁡n))O(n^3/2^{\Omega(\sqrt{\log n})}) time.

AI involvementAI-led
ImportanceNotable
Technical verificationFormal (Lean + human)
Human understandingFew
Paper date:

Sources

  1. Cui, Liu and Woodruff’s paper
  2. Section 7, AI disclosure
  3. Pinned Lean formalization notes
  4. Main theorem and proof of the two bounds
  5. Abboud and colleagues’ graph decomposition
Publication history