Every idea extracted from recent arXiv mathematics papers — verified and unverified. Click an idea to open its full card; badges show the empirical verdict.
Use an online estimate of the positive feedback gain among logits, routing probabilities, and representations to adjust the softmax temperature. Increase temperature when the estimated cyclic gain approaches the instability regime, preventing exponential amplification and router collapse without globally weakening all layers.
Add a weighted reflection symmetry to an attention or graph-propagation matrix instead of requiring ordinary permutation equivariance. For paired positions or graph nodes related by an involution, penalize the failure of the propagation operator to commute with the weighted reflection; this makes all geometric multi-step propagations symmetry-compatible. The method is suitable for data with mirror, reversal, paired-agent, or left/right structure where the two sides have unequal importance…
Build a kernel aggregation layer whose output is a tangent vector field on the unit sphere and whose surface divergence is identically zero by construction. For each source point, use a matrix kernel obtained by applying a surface-rotated gradient in the query variable to a scalar zonal kernel; this is a differential-form version of the paper's matrix-valued construction. The layer can replace attention or message passing when the target dynamics are incompressible, such as spherical fluid…
Penalize short positive feedback cycles in an iterative neural module by suppressing products of absolute Jacobian blocks around the cycle. This targets the mechanism responsible for exponential temperature sensitivity rather than merely penalizing the total Jacobian norm, allowing strong feed-forward paths while controlling recurrent amplification.
Use signed spanning-forest minor numerators to encourage a graph-structured neural layer to preserve independent multi-coordinate responses instead of collapsing several outputs onto the same direction. The determinant coefficients are only 0 or ±1, making the regularizer combinatorial and sign-exact rather than a noisy learned determinant surrogate.
Replace the linear phase progression in a positional encoding or rotary attention mechanism with a deterministic quadratic phase. The resulting position signal is generated by an irrational rotation with linearly changing increments, and the paper proves that its infinite diffraction measure is purely absolutely continuous, suggesting disorder-like spectral coverage without random sampling.
Replace independent top-k token selection by a quality-weighted determinantal subset objective based on the Hardy–Szegő kernel. Tokens with high learned quality are preferred, but geometrically redundant tokens have a small determinant contribution, encouraging diverse sets of routed experts, retrieved items, or attended context tokens.
Treat each attention head as a directed Markov graph and penalize token pairs that require many propagation steps to reach one another. This discourages isolated attention communities and slow information mixing while preserving the ordinary task objective.
Replace unconstrained predicted pairwise similarities with a correlation matrix whose diagonal is exactly one and whose spectrum is explicitly prevented from entering the nearly singular regime typical of high-dimensional elliptope samples. Add a soft spectral barrier during training and use a PSD-safe factorization at inference, so the model can represent dense correlations without relying on an unstable nearest-correlation-matrix repair.
Regularize a learned self-adjoint contraction so that its eigenvalues move toward 0 or 1 rather than accumulating in the transition interval. This suppresses ambiguous mixing modes and can enable a smaller binary spectral approximation at inference.
Regularize a neural attention or routing distribution according to how quickly it mixes toward a specified graph-dependent target, instead of penalizing only entropy or one-hop variation. The regularizer discourages pathological concentration on isolated graph regions while still allowing meaningful local structure, because concentration is judged after several graph-constrained Metropolis-Hastings steps.
Replace one independently learned vector per graph edge with a truncated spectral expansion on the line graph. The model learns coefficients for low-frequency edge modes and reconstructs edge features before message passing, reducing parameters while imposing an inductive bias that incident edges should have correlated behavior.
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.
Add a structured token-mixing layer based on commuting sums of swap operators rather than unconstrained pairwise attention. The layer learns a low-degree spectral filter in the Jucys–Murphy operators, allowing it to represent hierarchical interactions while retaining an explicit algebraic inductive bias.
Apply a convex Husimi functional as a differentiable regularizer to positive matrices used by attention heads, routers, or feature covariances. Penalizing the squared response suppresses sharp spherical peaks and can prevent collapsed routing or unstable attention without directly forcing uniform eigenvalues.
Regularize the eigenvalue spectrum of a neural representation or attention Gram matrix using the paper's universal-kernel spread-complexity curve. The loss penalizes spectral profiles that exhibit excessive level clustering or near-degeneracy, while allowing the desired amount of eigenvalue repulsion to be selected by a GOE-like, Poisson-like, or empirically calibrated target.
Add a positive multiplicative perturbation to the node or token measure of a symmetric neural operator and use the paper's eigenvalue-response matrix to identify nearly degenerate eigenspaces. Train the perturbation or its scale so that repeated eigenvalues split with a controlled minimum gap, making spectral positional encodings and eigenvector-based message passing more stable.
Add a clique-aware penalty to a learned graph adjacency or graph-attention matrix that suppresses excessive squared positive eigenvalue energy. Unlike a spectral-radius penalty, this controls the entire positive spectral subspace and can discourage highly concentrated, unstable message-passing channels while preserving useful negative-spectrum structure.
Replace an unconstrained bilinear feature interaction with a joint spectral filter that only allows pairs of graph or spherical frequencies satisfying a soft radius constraint. The smooth factor attenuates interactions near and beyond the cutoff instead of making the hard low-pass decision used by ordinary spectral truncation, which should reduce high-frequency aliasing and unstable feature products.
Add a spectral regularizer to a learned graph or sparse attention adjacency that penalizes violation of the paper's energy floor. The regularizer discourages adjacency matrices that retain many edges but collapse into a low-dimensional spectral structure, which may reduce graph-message-passing diversity and worsen oversmoothing.
Add a centro-affine Dirichlet penalty to a neural module whose inputs or outputs lie on a sphere, such as normalized embeddings or attention directions. The penalty measures intrinsic variation under an unconditional convex-body metric while projecting out the constant and coordinate-affine modes excluded by the theorem.