Every idea extracted from recent arXiv mathematics papers — verified and unverified. Click an idea to open its full card; badges show the empirical verdict.
Distill a teacher's attention into a student by matching sink mass and the normalized content distribution as separate targets rather than applying one KL divergence to the entire attention row. Use the Aitchison distance on the content composition, which compares relative token allocation and prevents a large common sink probability from overwhelming differences between content tokens.
Add a consistency loss for a neural model processing irregularly sampled functions or sequences, comparing its latent covariance representation before and after a time-coordinate warp. The target discrepancy should be near zero for affine transformations but should grow with the normalized curvature, bandwidth, and local sparsity of a nonlinear warp.
Use a small set of known or trusted pairwise costs to estimate the effective entropic temperature of a Sinkhorn attention or mixture-of-experts routing layer directly from its observed transport plan. This provides a calibration controller that can detect over-concentrated routing and adjust epsilon without backpropagating through a costly temperature search.
Construct training and evaluation examples whose every (d-1)-variable marginal is exactly independent, but whose full d-variable distribution contains a parity interaction. This isolates genuine high-order reasoning from shortcuts based on pairwise or lower-order statistics and can expose whether attention or MLP architectures learn the intended interaction.
Replace raw pairwise attention or graph-edge scores by exact U-centered residuals, removing additive effects attributable to either endpoint. The resulting scores represent interaction beyond independent source and destination biases and satisfy zero row sums, preventing a few high-degree or high-activation tokens from dominating relational aggregation.
Add a degree-conditioned neighborhood-profile penalty to a GNN so that its effective message-passing graph has a controlled hub-neighborhood trend. The regularizer can either target a rank-one null profile, where neighbor degree is approximately independent of root degree, or deliberately target a learned/reference logarithmic trend when preferential-attachment-like structure is useful.
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.
Add an auxiliary loss that makes selected representation coordinates insensitive to all subsets of fewer than d variables while retaining a d-way parity statistic. The objective discourages the network from solving a task through pairwise shortcuts and explicitly rewards a controlled high-order interaction.
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 centered triangle-consistency term to a graph neural network or graph transformer. The term rewards learned edge affinities whose triangle products exceed the independent-edge baseline while preserving the overall edge density, encouraging locally coherent neighborhoods instead of arbitrary pairwise affinities.
Build an attention or positive-mixture module whose output ratio at two control settings is provably monotone in an ordered index such as token distance, retrieval rank, or discretized uncertainty. Use normalized-positive-series identities to replace an unstable quotient derivative with a difference of expectations, and penalize violations of the resulting stochastic-order condition during training.
Calibrate the maximum attention logit in each head against the log-correlated extreme-value law instead of applying fixed clipping or a fixed max-norm penalty. Penalize only maxima that exceed the predicted log N minus three-quarter log log N baseline by an unusually large order-one fluctuation, allowing ordinary sharp attention while suppressing rare pathological spikes.