New statistical methods for analyzing data where observations are grouped (like patients in hospitals) and each group provides multiple related measurements.
Anytime-Valid Inference in Linear Models with Clustered Outcomes: Exact Multivariate Tests and Cluster-Robust Confidence Sequences · Michael Lindon · arXiv:2610.25613
A new mathematical bridge, inspired by the famous Atiyah-Segal homomorphism, to connect "approximate representations" of infinite groups with topological spaces. These approximate representations are functions that behave almost, but not quite, like perfect representations.
An Atiyah-Segal homomorphism for approximate representations of infinite groups, and almost multiplicative ucp maps · Rufus Willett · arXiv:2610.24252
This paper determines the absolute minimum number of computations required to solve a common type of well-behaved optimization problem on hyperbolic space, a curved geometry gaining traction in fields like machine learning. It proves that an existing optimization method is already perfectly efficient, meaning no algorithm can do better.
Tight lower bounds for Lipschitz geodesically convex optimization on hyperbolic space · Christopher Criscitiello · arXiv:2610.26815
New mathematical tools to analyze how 4-dimensional spaces can be continuously reshaped. It uncovers an amount of hidden complexity, showing that many transformations previously thought to be simple or "trivial" are actually distinct and intricate.
Topological mapping class groups of non-simply connected 4-manifolds · Jae Choon Cha · arXiv:2610.26400
This paper simplifies the underlying math for a powerful algorithm that quickly and accurately estimates how many ways you can perfectly pair up items in a complex network. It makes a key part of OpenAI's original proof much more direct and understandable, removing complicated auxiliary steps while keeping the algorithm just as robust and effective.
A simplified proof of the FPRAS for counting perfect matchings · Chenghua Liu, Boning Meng · arXiv:2610.26803
A general method to count how many different ways you can break down large, complex networks (hypergraphs) into many identical smaller pieces. It then applies this technique to determine the asymptotic number of "universal cycles," which are special sequences that list every possible short combination of elements exactly once.
Asymptotic Enumeration of $F$-decompositions in Typical Hypergraphs and of Universal Cycles · Danni Li, Bin Wang, Junxue Zhang · arXiv:2610.26579
A new computational method to analyze the complex 3D spaces associated with knots, allowing researchers to determine their "unknotting number", the minimum number of times you must untangle a knot to make it a simple loop.
Negative-Definite Plumbings and New Unknotting Numbers · Seong-Jin Lee · arXiv:2610.25430
This paper disproves Arveson's hyperrigidity conjecture, a significant open problem in operator theory. A specific "boundary" condition doesn't always guarantee a stronger "rigidity" property for mathematical structures called operator systems, even in simpler, well-understood settings.
Commutative Counterexamples to Arveson's Hyperrigidity Conjecture · Boris Bilich, Marcel Scherer · arXiv:2610.26556
A look at how a specific, somewhat technical, measure of how space curves can determine its overall shape. If a space satisfies a condition called "103-positive curvature operator of the second kind," it must be topologically equivalent to a sphere or a related "spherical space form."
Sphere Theorems for $\frac{10}{3}$-Positive Curvature Operator of the Second Kind · Haiqing Cheng, Kui Wang · arXiv:2610.26269
This paper rigorously studies the shapes that a flexible, elastic loop (a knot) takes when it tries to minimize its bending, particularly focusing on what happens when parts of the knot touch each other.
Elastic knots: compactness, contact, and variational structure · Philipp Reiter, Heiko von der Mosel · arXiv:2610.26823