Meta AI Research published six papers produced through collaborations between mathematicians and Muse Spark. The company says five of the papers present answers to previously open research questions.
The researchers used Muse Spark 1.1 and 1.2 in Thinking Mode through the regular Meta AI chat interface, without a custom research scaffold. Meta describes the goal as helping researchers develop mathematical insights, rather than producing papers at scale.
Researchers guided and checked the work
Meta says mathematicians selected and guided the research, working with Muse Spark to explore ideas and develop arguments. A second group of mathematicians reviewed the work. Each paper identifies passages drafted primarily by researchers and passages drafted primarily by AI, according to the announcement.
The reported division of work varied by project. For a group-theory paper, Muse Spark generated a GAP search program that found a counterexample, which researchers verified and developed into an argument. In the arithmetic-physics project, Meta says the model helped connect ideas from number theory and p-adic string theory, generated candidate proofs and drafted three technical sections that researchers checked, corrected and refined.
Six areas of mathematics
The papers cover probability, differential equations, group theory, optimization, arithmetic physics and non-associative algebra. Their topics include a threshold for fitting Gaussian points to an ellipsoid, wave collapse, counterexamples to algebraic conjectures and a connection between two-point functions and height functions on curves.
Meta says it learned after completing the work that outside teams had independently announced solutions to some of the same problems using different approaches. The company says those contributions and their relationship to its papers are acknowledged in the papers.
