Meta's Muse Spark Model Helps Mathematicians Solve Six Open Problems in Six Months

Bi Shuchao, head of Meta's Superintelligence Lab, announced on X on October 5, 2026, that the lab's Muse Spark 1.1 and 1.2 models had assisted leading mathematicians in publishing six papers across six fields—probability theory, differential equations, group theory, optimization, arithmetic physics, and non-associative algebra—over the past six months. Five of the papers provided answers to open problems, and one refuted the Kida conjecture in group theory by identifying a semiabelian non-monomial group of order 384.

According to Bi, Muse Spark generated complex search code to find the counterexample in a mathematical software system. In other results, the model helped process large-scale computations for the mass-critical biharmonic nonlinear Schrödinger equation, derived a strict threshold for Gaussian ellipsoid fitting, linked p-adic string theory with number theory, and drafted three core technical sections of a paper. The model also contributed key conditions and counterexamples in optimization theory and evolution algebras.

Bi said the mathematicians used only the web interface of meta.ai with Muse Spark's thinking mode and natural-language prompts. The six papers have passed independent double review, with human- and AI-written parts clearly marked. Bi predicted that similar breakthroughs could expand into biology, chemistry, and physics.