Which open problems did Muse Spark help solve?

Meta's six Muse Spark papers include a sharp threshold for fitting random Gaussian points to ellipsoids, finite-time blow-up for the mass-critical biharmonic nonlinear Schrödinger equation (open since 2015), a counterexample showing semiabelian groups need not be monomial, exactness of a cycle-based optimization relaxation, and a disproof of a conjecture about solvable evolution algebras.

Did Meta use a special research system for the Muse Spark math papers?

No. Meta says the mathematicians used Muse Spark 1.1 and 1.2 in Thinking Mode through the ordinary meta.ai chat interface, with no custom research scaffold. The human researchers guided the work and kept final decisions, and a separate group of mathematicians reviewed each paper before publication.

Were these Muse Spark results found first?

Not always. Meta's post notes that three independent solutions to the Gaussian ellipsoid fitting problem appeared in August 2026, an AI agent called Nilradical reported a different semiabelian-group counterexample on 16 September 2026, and Hu and Wen reported evolution-algebra counterexamples at the same time.

How do the Muse Spark papers show what the AI wrote?

Each of Meta's six papers marks which sections Muse Spark drafted and which the researchers drafted, and the paper pages include a statement of AI use. In the string theory paper, for example, Muse Spark generated candidate proofs and drafted three core technical sections.

Meta · 2026-10-02 · major

Muse Spark math papers — Meta's model helps settle five open problems

Meta published six math papers written by mathematicians working with Muse Spark 1.1 and 1.2 in Thinking Mode in the normal Meta AI chat. Five answer open questions; separate mathematicians reviewed each one.

Hero graphic for Meta's Solving Open Research Problems Together post on Muse Spark math papers

Mathematicians used Meta's Muse Spark in an ordinary chat window to answer five open problems across six papers.

Quick facts

Maker	Meta (with outside mathematicians)
Models	Muse Spark 1.1 and 1.2, Thinking Mode
Interface	Standard meta.ai chat, no custom scaffold
Papers	6 (5 answer open problems)
Fields	Probability, PDEs, group theory, optimization, algebra, arithmetic physics
Published	2 October 2026

What is it?

Meta's 'Solving Open Research Problems Together' post publishes six mathematics papers co-written by human researchers and Muse Spark 1.1 and 1.2. Five of them answer questions that were previously open, in probability, differential equations, group theory, optimization and non-associative algebra; the sixth links p-adic string theory and number theory.

How does it work?

The mathematicians used Muse Spark in Thinking Mode through the regular meta.ai chat, with no custom research scaffold. They guided the work, a separate team of mathematicians reviewed each paper, and every paper marks which sections were drafted by the AI and which by people. In one example, Muse Spark found a 384-element group that disproves a 2024 conjecture about semiabelian groups.

Why does it matter?

Results like these usually come from labs with special agent systems; here they came from a consumer chat model. Meta is also open about the limits: several problems were solved at the same time by other people or AI agents, which shows how crowded AI-assisted math has become.

Who is it for?

mathematicians and AI-for-science researchers

Frequently asked questions

Which open problems did Muse Spark help solve?: Meta's six Muse Spark papers include a sharp threshold for fitting random Gaussian points to ellipsoids, finite-time blow-up for the mass-critical biharmonic nonlinear Schrödinger equation (open since 2015), a counterexample showing semiabelian groups need not be monomial, exactness of a cycle-based optimization relaxation, and a disproof of a conjecture about solvable evolution algebras.
Did Meta use a special research system for the Muse Spark math papers?: No. Meta says the mathematicians used Muse Spark 1.1 and 1.2 in Thinking Mode through the ordinary meta.ai chat interface, with no custom research scaffold. The human researchers guided the work and kept final decisions, and a separate group of mathematicians reviewed each paper before publication.
Were these Muse Spark results found first?: Not always. Meta's post notes that three independent solutions to the Gaussian ellipsoid fitting problem appeared in August 2026, an AI agent called Nilradical reported a different semiabelian-group counterexample on 16 September 2026, and Hu and Wen reported evolution-algebra counterexamples at the same time.
How do the Muse Spark papers show what the AI wrote?: Each of Meta's six papers marks which sections Muse Spark drafted and which the researchers drafted, and the paper pages include a statement of AI use. In the string theory paper, for example, Muse Spark generated candidate proofs and drafted three core technical sections.