Computer-assisted Proof Under Audit: Typos, Certificate Errors, and Reproducible Exact Checks for a Symbolic Invertibility Proof
This paper presents the first independent, source-level audit of a published computer-assisted proof in analysis, revealing 11 proof-affecting defects in the original certificate that invalidate the claimed conclusion despite the underlying theorem remaining potentially true.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the universe as a giant, churning ocean of invisible fluids. Sometimes, these fluids get so excited that they try to fold in on themselves, creating a "singularity"—a point where the math breaks down and the rules of physics seem to vanish. Scientists are obsessed with figuring out exactly how and why this happens, because understanding these cosmic crashes helps us predict everything from weather patterns to the behavior of stars. To solve these puzzles, mathematicians often build complex models, like intricate LEGO castles, to prove that a specific part of the fluid will behave in a certain way. But here's the catch: when the castles get too big to build by hand, scientists ask computers to help. They write code to check the math, hoping the machine will spot the tiny cracks in the foundation that a human eye might miss. This is called a "computer-assisted proof," and it's like handing a robot a magnifying glass to inspect a billion tiny bricks.
But what happens if the robot is looking at the wrong bricks, or if the instructions it was given have a few typos? That is the story of this paper. A researcher named Fan Zheng decided to act as a "mathematical auditor" for a very famous, recently published proof about these fluid singularities. The original paper claimed to have proven that a specific mathematical tool (an operator) could be "inverted"—a fancy way of saying it could be reversed to solve the puzzle—using a computer to do the heavy lifting. Zheng didn't just run the code again; they went deep into the source code and the printed formulas, checking every single step like a detective looking for clues.
The audit found that while the original idea was likely still a good one, the "certificate" (the computer-generated proof) was broken. Zheng discovered 11 specific defects that meant the computer's proof didn't actually prove what the author claimed. It wasn't that the whole theory was wrong, but rather that the specific evidence presented was flawed. The paper found things like missing pieces in a puzzle, signs that were flipped upside down, and numbers that were slightly off. The authors of the original paper had published a corrected version in a top journal, but the auditor found that even the new version still had the same mistakes in the code and formulas. The paper concludes that the original computer-assisted proof is not rigorous yet; it needs to be rebuilt with a cleaner, simpler design to truly work. It's a reminder that even when a computer says "I did it," we still need a human to double-check that it actually did the right thing.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.