Meta's Muse Spark co-authors six mathematics papers
Meta mathematicians wrote six research papers with Muse Spark, five answering previously open questions, using the ordinary chat interface rather than a custom research system.
What happened
On October 2 Meta said a team of its mathematicians wrote six research papers with its Muse Spark model, five of which answer previously open questions. AI-drafted and human-written passages are marked, and a second group of mathematicians reviewed the work.
The problems
The papers cover probability (the strict threshold for Gaussian ellipsoid fitting), differential equations (finite-time blow-up for a biharmonic nonlinear Schrödinger equation), group theory (semiabelian groups need not be monomial), optimization, arithmetic physics and non-associative algebra, including a counterexample on solvable evolution algebras and a disproof of a 2024 conjecture.
How it was done
The researchers used Muse Spark 1.1 and 1.2 through Meta AI's regular chat interface, with no custom research scaffold. Similar results from DeepMind and OpenAI have usually relied on specialized systems.
Why it matters
Math results are verifiable, so marked contributions plus independent review give a transparent answer to whether the AI really did the work, and a model for crediting AI in science.
What an open problem is
An open problem is a statement not yet proven or disproven. A few are famous; most are narrower questions researchers flag in papers. Muse Spark's contributions are of the second kind, but closing them with peer-reviewed results is how science normally advances.
Limits
Humans chose the ideas, steered the problems and verified results. The model acted as a research partner, and the solved problems are not the field's most famous ones.
Competitive context
Meta is using Muse Spark and the Muse app to answer criticism that it trails OpenAI, Google and Anthropic. Verifiable math results showcase reasoning, while the claim that Muse read private messages on Mac shows privacy questions remain.
Why it is credible
Marking which passages the model wrote and sending the work to a second group of mathematicians gives outsiders a way to check the claims, unlike benchmark scores published without review.
What it means for your business
General chat tools can now be real research partners for experts. Pair your specialists with AI, record which parts the AI produced, and build independent verification into the process.