Every idea extracted from recent arXiv mathematics papers — verified and unverified. Click an idea to open its full card; badges show the empirical verdict.
Add a profile-preserving exchange layer that mixes representations through pairwise state swaps, with weights shared over all positions. Unlike ordinary token-wise MLP mixing, the layer explicitly models interactions between two categorical states while respecting the symmetry of position permutations.
When a chosen sparse support is geometrically incompatible with exact orthogonality, temporarily optimize on a nearby off-diagonally perturbed Stiefel constraint rather than forcing a singular Newton system. Anneal the perturbation to zero after the active support has stabilized, using the paper's O(||Delta||_F) KKT guarantee to control the residual of the original orthogonality-constrained problem.
Replace explicit quotient construction by a differentiable projection that removes learned group-orbit directions from both source and target features. The paper's reduction argument shows that a closed level constraint makes the restricted form horizontal, so the network can operate on invariant coordinates while retaining a measurable residual for symmetry leakage.
Replace independent top-k MoE routing with a submodular polyhedral allocation over experts. A learned set function assigns a marginal gain to each additional expert allocation, so the router exhibits diminishing returns and can enforce global capacity constraints rather than making unrelated per-token choices. The allocation is obtained by sorting marginal gains, giving a fast greedy router with piecewise-linear routing regions.
Add a coordinate-aware long-range aggregation branch whose singular low-frequency component is explicitly centered before it is mixed into token representations. The centering acts as a neural counterterm: constant or slowly varying value fields cannot accumulate an activation contribution that grows with context size, while local and higher-frequency interactions remain available through an ordinary attention residual branch.
Add a targeted barrier or hinge loss to an existing attention or graph-mixing matrix that penalizes violations of signed circular-minor inequalities. Instead of enforcing only generic entrywise positivity, constrain higher-order noncrossing interactions encoded by determinants. This can suppress pathological oscillatory mixing while still allowing individual entries to be negative when the global structured sign pattern permits them.
Augment each token or graph node with a periodic latent position x_i and phase θ_i, then evolve these variables before attention or message passing. Tokens with similar phase attract in x, while tokens with similar position synchronize in θ, producing self-organized groups without an externally specified clustering objective. The coupling strengths J and K provide interpretable controls for aggregation and synchronization, and their sweep should expose the paper's four collective regimes and…
Replace a standard permutation-invariant object pool with a latent state on an unordered configuration together with a fiber vector transported along the observed object trajectories. The instantaneous state remains invariant to reordering, but loops and exchanges of objects act through learned monodromy matrices, allowing the network to represent path-dependent interactions without assigning arbitrary permanent object indices.
Represent a large positive semidefinite neural operator as the sum of two Kronecker products and regularize an efficiently computed upper bound on its largest eigenvalues. The bound controls not only the spectral norm but every top-k eigenvalue sum, allowing a tunable penalty on concentrated or unstable directions without constructing the exponentially larger operator.
Replace iid dropout or iid activation noise on spatial tokens with fluctuations generated by a conserved diffusing density. Each token receives a positive mass variable whose total mass is preserved, while Poissonian stochastic flux produces correlated perturbations that explore coherent local patterns rather than independently corrupting every feature. The density is autonomous and detached from autograd, so the regularizer adds little computational overhead.
Add a training-time regularizer that keeps the empirical joint covariance of hidden activations on multiple inputs close to the recursively predicted NNGP covariance. The regularizer targets the finite-width fluctuations quantified by the Wasserstein result, and is particularly appropriate for recurrent networks and attention blocks with shared weights, where hidden states at different positions or time steps are statistically coupled.
Replace coordinate-wise mean pooling of metric-valued items with a finite representation of their free integral. Each item x in a pointed metric space M is represented through evaluations of learned Lipschitz probes, and the pooled feature is the weighted integral of those probe values. A dual Lipschitz critic estimates the free-space norm of differences between pooled groups, making the representation sensitive to metric geometry while remaining permutation-invariant.
Add a learned Riesz-transform branch that extracts normalized spatial gradients after diffusion by a positive parabolic operator. The diffusion branch carries smooth semantic content, while the Riesz branch represents boundaries, motion changes, and graph discontinuities. Resolvent smoothing makes the derivative branch less sensitive to feature noise than directly applying a finite difference.
Construct a sparse attention or message-passing graph as a gain graph, where every directed edge carries a relative frame label and reverse edges carry the inverse label. Grow the graph using the paper's 2-extension operation: replace two old edges by a new vertex connected to their four endpoints, while preserving the relative gain products. The resulting mask is intended to preserve global information flow under controlled sparsity and to avoid isolated components and brittle bridges commonly…
Replace a dense channel or token-mixing matrix with a product of positive bidiagonal factors, so information propagates through a controlled sequence of local couplings rather than arbitrary signed interactions. Initialize the factors from the paper's barycentric-subdivision factorization, then learn positive diagonal and off-diagonal parameters; the resulting map is structured, parameter-efficient, and constrained to remain totally positive.
Replace a free-form order-dependent gate with a positive mixture of Mellin powers $(1+s)^{-a}$. This gives a small, interpretable module whose response across the order variable is automatically generated by a positive measure and therefore inherits complete monotonicity, log-convexity, and Hankel-moment structure.
Parameterize a relative-position or lag-decay function as a finite positive mixture of exponentials instead of learning arbitrary attention bias values. The resulting kernel is completely monotone on positive distances, so it is nonnegative, decreasing, and has alternating derivative signs; the mixture provides several learned memory scales without allowing oscillatory or unstable long-range biases.
Use a learned asymmetric Finsler-like cost instead of the symmetric Euclidean distance in attention logits. The metric has a Riemannian quadratic part and a directional drift term, while a differentiable barrier enforces the strong-convexity condition derived for the paper's extended $(\alpha,\beta)$-metrics. This lets each attention head prefer one direction in feature space without producing pathological, non-convex distance landscapes.
Replace ordinary additive path aggregation in graph attention with ordered products of edge operators equipped with learned reversal and color-switch maps. Closed-loop products become a consistency signal, allowing the model to retain direction-sensitive relational information that standard permutation-invariant message passing can lose.
Add a pseudo-determinant-based connectivity objective to a neural model that predicts graph edge weights, attention adjacency, or sparse routing links. Maximizing the Laplacian pseudo-determinant rewards many globally distributed spanning trees, discouraging disconnected or bottlenecked learned graphs without requiring a discrete connectivity constraint.
Use the paper's q-ary overlap inequality as a regularizer for categorical neural networks. Two independently sampled attention, routing, or message-passing supports should rarely overlap in many locations; penalizing the moment q^{|S\cap S'|} discourages redundant histories and correlated interference between heads or experts.
Estimate the temporal spectrum of each sequence channel using a locally private procedure, then apply a regularized inverse-square-root spectral filter before the sequence enters attention or an SSM. The filter removes predictable low-frequency or narrow-band redundancy while avoiding unstable amplification at frequencies where the private estimate is small.
Replace unconstrained combinations of several attention or adapter operations with a brace-style composition that inserts each operation into a distinct ordered interval of a base sequence. The resulting computation preserves the order of host and inserted operations and forbids crossing dependencies, producing hierarchical attention patterns with an explicit structural bias.
Represent each of m neural branches by a positive input field f_i and a positive output field g_i, then penalize violations of the paper's multi-output Borell-Brascamp-Lieb bound at weighted barycenters. The constraint couples branches through both local normalized ratios and global mass ratios, encouraging calibrated multi-view predictions without requiring all output functions to be identical.