Meta claims AI-assisted answers to five mathematics research problems

Meta claims AI-assisted answers to five mathematics research problems

Image: Primary

Published Sunday, October 4, 2026 · 1:18 AM EDT

Meta released six papers on October 2 co-authored by mathematicians and its Muse Spark AI, claiming that five answer previously unsolved research problems. The researchers used Muse Spark 1.1 and 1.2 through the public meta.ai chat interface. In one paper, the AI wrote a search program for GAP mathematical software that found a group with 384 elements, providing a counterexample to the Kida conjecture. Mathematicians selected and led the research, and a separate group reviewed the results. The novelty is contested: researcher Jason Dean Lee said three problems had already been solved. Meta also acknowledged independently obtained results for three papers.

Sources

In this story

Published by Tech & Business, a media brand covering technology and business. This story was sourced from note and reviewed by the T&B editorial agent team.

Back to Newswire

Keep reading

Products Policy

Products Policy

Boeing awarded US$20-billion Navy fighter development contract

Boeing has been awarded a US$20-billion US Navy contract to develop the F/A-XX sixth-generation fighter, New Atlas reports. Boeing beat rival Northrop Grumman for the contract. The Navy needs a multirole aircraft that can operate...

Products

Products

Valley National agrees to buy Bluevine for $340M

Valley National Bancorp has agreed to acquire New Jersey-based Bluevine for $340M, CTech reports. Bluevine provides digital banking and payments services to 175K small and medium-sized businesses in the US. The acquisition would ...

AI Science

AI Science

Meta says Muse Spark wrote search code that disproved a symmetry conjecture

Meta says researchers disproved a proposed rule about mathematical structures that describe symmetry using a counterexample found by search code generated by Muse Spark. The research team verified the result and completed the proo...

AI Science

AI Science

FP4 pretraining study reports faster throughput with training-loss tradeoffs

Researchers report in an arXiv preprint that a custom four-bit floating-point training route reached 37.9K tokens/s/GPU, compared with 18.8K for bfloat16 and 27.6K for Transformer Engine in matched tests on the same accelerator. T...

AI Science

AI Science

Nous preprint demonstrates memory improvement certificates without source calibration

A preprint presents a method for certifying improvements in an agent's memory decisions without first estimating how reliable its information sources are. In one mathematical model family, learning and certifying useful decisions ...

AI Science

AI Science

Researchers use brain signals to distinguish and correct AI errors in simulations

Researchers at KAIST and Microsoft Research Asia demonstrated a system that uses brain signals to distinguish whether an AI has misunderstood a person's goal or chosen an unexpected method. Their Neural Value Alignment technique f...