OpenAI Announces Three Results in Mathematics, Computer Science, and Physics; Academia Questions Release Method

OpenAI Announces Three Major Scientific Breakthroughs, but the Research Community Is Displeased With How It Released Them IT Home Download the Client | Exclusive First Look October 28, 2026, 11:27:59 From Shandong Updated October 28, 2026, 11:34:07

IT Home, October 28 — OpenAI announced on Monday local time that the company had achieved remarkable breakthroughs in mathematics, computer science, and physics, while saying it is adjusting the way it communicates about releasing its results. But despite what looks like a course correction, researchers interviewed by The Verge are growing increasingly weary of how the company operates.

IT Home notes that over the past few months there has been constant friction between OpenAI and the mathematics community. In early October, the company dumped hundreds of preprints at once, leaving mathematicians struggling to keep up. This time, OpenAI said it would work with domain experts serving as "guides" to determine the appropriate way to release its latest results.

The three results each target some of the most important unsolved problems in mathematics, computer science, and physics; some continue OpenAI's previous lines of research, while others open new research directions. Based on the advice of these "guides," the three results used different release approaches: for the mathematics result, the AI-generated derivation content was released directly, "leaving it to academic researchers to organize themselves to interpret the conclusion"; for the computer science result, OpenAI will form an external research group to handle interpretation; and the physics result will not be made public for now, with OpenAI saying it will leave it to an external expert team to decide the subsequent release format.

The mathematics result is stunning: it is said to have proven the Kakeya conjecture in all dimensions. Simply put, this problem studies how much space is minimally needed to allow a needle or a pencil to rotate to point in every direction. Earlier this year, mathematician Wang Hong won the prestigious Fields Medal precisely on the strength of results including the proof of the three-dimensional version of the conjecture. Brown University mathematics professor Brendan Hassett said OpenAI's conclusion "looks well-founded," but stressed that he had not yet verified the Lean formal proof, the formal system that relies on computers to check conclusions.

The physics result centers on the Standard Model, the most mature and complete theoretical framework in physics. The research focuses on quantum chromodynamics, the theory describing the strong nuclear force, one of the four fundamental forces, and its connection to Yang-Mills theory; Yang-Mills theory is also a cornerstone of modern particle physics and one of the seven Millennium Prize Problems, though OpenAI's research did not resolve the outstanding "mass gap" problem.

The final result claims to disprove the exponential time hypothesis. This is a widely accepted assumption in theoretical computer science: that some problems cannot be solved by methods much faster than brute-force enumeration. The hypothesis is closely related to the P versus NP problem, which is also a Millennium Prize Problem and one of the best-known unsolved problems in the field; its core question is whether problems whose answers are easy to verify can themselves also be solved quickly.

Because the results vary in how publicly they have been released, and because verification thresholds for such results are extremely high, researchers find it difficult to fully assess these conclusions. Even with all the materials, verification would require a great deal of time. But the experts interviewed said the results deserve serious attention, and many described them as "enormous in weight," "major breakthroughs," and "far-reaching."

Researchers are certainly excited by these scientific advances, but they are also annoyed by OpenAI's approach. After months of receiving a flood of AI-produced research conclusions, this fatigue is already very familiar. Many researchers hope the entire academic community can concentrate its efforts on the arduous verification work, and they question why OpenAI would add further barriers by hand-picking experts to control the entire process.

Most experts interviewed approved of OpenAI seeking input from external researchers, but they argued that publicly announcing research conclusions means the relevant materials should be opened to everyone for verification. Constantin Kogler, a researcher at the Institute for Advanced Study, said: "Compared with dumping hundreds of papers at once before, releasing only three results this time — I approve of that pace." But he did not agree with limiting access to experts chosen by OpenAI. He said: "There is no reason to allow only a few well-known scientists to see the content at the outset."

Andreas Karch, a physics professor at the University of Texas at Austin, said: "Having them hand-pick the expert team is a very bad thing. Large laboratories already have enough say; they should not be the ones deciding who is qualified to participate and who is excluded." Karch initially suggested selecting researchers through a selection mechanism, though he acknowledged that those ultimately chosen would most likely still be top scholars. He then asked: why not simply release the supporting materials at the same time, along with an expert interpretation group? "That way everyone would have a chance to access the content, and it would also reflect their desire for the results to be understood and taken seriously."

Just one day before OpenAI's official announcement, the Advisory Group on Mathematics and Artificial Intelligence (AGMAI) had just published an article specifically warning about this kind of information control behavior. The expert group stated: "We do not recommend that AI companies adopt this model." They believe that prioritizing access to content for specific groups is unnecessary; this practice borrows the prestige of the selected researchers and also brings a series of risks, including unequal access, biased selection, and unexplained delays in progress. The expert group's proposal is: AI labs should upload their results to academic preprint platforms, along with formal proofs and explanations of their topic selection, while also providing funding and supporting resources to help the academic community interpret these conclusions.

After this batch of new results was published, AGMAI urged OpenAI to "immediately publicly release the natural-language manuscript and Lean formal proof corresponding to this physics result":

These are both core questions in physics and important mathematical problems. The existence of quantum Yang-Mills theory is one of the Millennium Prize Problems, and the construction of four-dimensional interacting quantum field theory has been a goal in mathematical physics for decades.

We deeply respect the scientists whom OpenAI invited to lead this work. But as we stated previously, interpretation of the results should proceed in a public environment, with everyone having equal access. OpenAI said the delay in release was to complete the Lean formal proof. By their own account, that work is now complete, and from a scientific standpoint there is no reason to further delay publication.

Hassett believes that fully publicizing the results would also help advance scientific development and help new tools such as AI gain broader acceptance. "Multiple research teams around the world studying these conclusions together can enable researchers who need them to get started with these new tools more quickly."

Finding a way to communicate that satisfies researchers across all the disciplines OpenAI is entering may itself be unrealistic. OpenAI says it is trying different approaches and acknowledges that each approach involves trade-offs. Stanford University mathematician Jared Duker Lichtman said: "My feeling is that no matter which approach is chosen, there will always be someone who is dissatisfied. We just need to find the least controversial way."

Xiaoyu He, a professor of mathematics at Georgia Institute of Technology, believes the problem goes far deeper than the format of release. "The results themselves are very impressive, but in my view, they are merely deliberately projecting an attitude of responsibility and caution while avoiding the core contradiction: these foundational difficult problems were rapidly solved by AI, with no contribution from human researchers throughout the process, and the resulting proof texts are very difficult to read and hard to verify."

"Special statement: The above content (including videos, pictures, and audio if any) is uploaded and published by users of 'Dafeng Hao,' a self-media platform under Phoenix Net. This platform only provides information storage space services.

Notice: The content above (including the videos, pictures and audios if any) is uploaded and posted by the user of Dafeng Hao, which is a social media platform and merely provides information storage space services."