2581 valuable arXiv mathematics papers the pipeline has analyzed — each with a short note on what its math gives to machine learning.
The paper gives a directly actionable fairness principle for continuous LLM batching: because decode-step cost is governed by the largest active KV-cache footprint, co-batched requests should have bounded disparity in decode progress or…
The paper offers a strong orbit-separation mechanism for flows with singularities: Komuro expansivity remains valid when trajectories are compared under arbitrary increasing time reparametrizations. It also proves a sharp periodic-orbit…
The paper provides a constructive frequency-map mechanism: weighted Birkhoff averages recover frequencies of analytic quasi-periodic trajectories with exponentially small error rather than only polynomial error, under Diophantine or Brjuno…
The paper reduces a geometric variance objective to a finite configuration of branch points on the sphere and to log-partition integrals of exponentials of a Green-function potential. The transferable asset is a differentiable, globally…
The paper gives a constructive way to represent differential fields on a manifold with one local field per chart, connected by explicit transition laws rather than forcing a single globally valid coordinate representation. Its transferable…
The paper identifies a useful variance-reduction principle for black-box optimization under input uncertainty: retain the realized phenotype or input and Rao-Blackwellize the evolutionary-strategy score estimator by replacing a latent…
The paper supplies a useful geometric decomposition of a Transformer residual update into radial magnitude change and tangential direction change on the unit sphere. Its key transferable asset is that only the projected FFN field can alter…
The paper gives a strong finite-dimensional structure for Ollivier–Ricci curvature as a function of random-walk idleness: for every adjacent pair, the curvature is concave and has at most three affine regimes. This means a…
The paper’s transferable asset is an efficient realization of fractional powers of a graph Laplacian without eigendecomposition: rational approximations turn a nonlocal spectral operator into a small number of sparse shifted elliptic…
The paper develops integer-preserving local rewiring moves for traversing graph fibers with fixed vertex-degree and Forman-curvature statistics. The transferable asset is not curvature itself, but the construction of small, exactly…
The paper supplies a smooth spectral replacement for the matrix polar factor, preserving the singular-vector geometry of Muon while removing its singularity at rank-deficient momentum matrices. The key transferable asset is the bounded…
The paper develops optimality-preserving screening for cardinality-constrained convex quadratic programs by combining a perspective relaxation, Fenchel duality, and primal upper and lower bounds. The transferable asset is the ability to…
The paper identifies a noncommutative analogue of the Kalman rank condition: controllability is determined by the span of matrix words formed from multiple coupling operators applied to the control matrix. This reveals invariant…
The paper develops a Lovász-type geometric representation for signed graphs using unit vectors and an orthogonal involution, whose fixed and anti-fixed subspaces encode switching behavior. This structure transfers naturally to signed…
The paper provides a nonstandard input-output boundedness mechanism for positive feedback interconnections: a plant with persistent-input/persistent-output (PIPO) behavior amplifies sustained large inputs, while a controller with…
The paper gives an exact combinatorial certificate for when a pure simplicial complex has an extremal up-Laplacian spectrum: the difference between the squared conjugate vertex-degree partition and the squared nonzero spectrum equals twice…
The paper identifies a precise boundary kernel between short memory and power-law long memory: a_j = exp(-lambda j) ell(j)/j, whose cumulative mass grows logarithmically through L(N) = sum_{j<=N} ell(j)/j. This suggests a causal sequence…
The paper gives a long-horizon Gaussian decomposition for approximate message passing (AMP) started from an uninformative Gaussian vector, showing that an initially vanishing signal overlap can grow to weak recovery. The transferable asset…
The paper gives a concrete example of information that survives in a geometric and holomorphic representation but is annihilated by ordinary topological stabilization: all flat line bundles associated with surface-group characters have the…
The paper replaces ordinary monotonicity of an operator F with monotonicity of a pair (F,v), where v can be selected to make a difficult nonmonotone problem tractable and may have an easier inverse. This suggests a preconditioned implicit…
The paper offers a constructive mechanism for discovering continuous symmetries of stochastic dynamics: learn vector-field generators from trajectory increments while enforcing SDE determining equations, Lie-bracket closure, Lie-algebra…
The paper provides a constructive regularity mechanism for the top Lyapunov exponent of a Markov-switching quasi-periodic linear system: under a primitive transition matrix, non-resonant base dynamics, and a simple top exponent, the…
The paper provides a concrete observability perspective for finite-dimensional distribution matching: a sampled interaction field is a linear measurement \(\operatorname{vec}(V_X)=Mc\) of the antisymmetric coefficient mismatch, so small…
The paper provides a dimension-free anti-concentration principle for vector-valued polynomial pushforwards: an absolutely continuous polynomial image of a log-concave distribution cannot place too much mass in a small output set. Its most…