Every idea extracted from recent arXiv mathematics papers — verified and unverified. Click an idea to open its full card; badges show the empirical verdict.
Cluster recurrent modules or MoE experts by the geometry of their observed finite-horizon input-output behaviors rather than by parameter distance. Train one shared optimizer/controller or low-rank adapter per cluster while retaining module-specific parameters and routing. This should reduce control and optimizer overhead without merging modules whose temporal responses are dynamically incompatible.
Replace Euclidean projection onto a smooth convex parameter constraint by a local tangent-halfspace linear minimization step. The method is particularly suitable for convex softmax heads or low-dimensional adapter parameters constrained to an ellipsoid, where projections require matrix solves or expensive nonlinear routines but membership and boundary-normal queries are cheap.
Replace repeated multi-task training runs at different loss weights with pseudo-arclength continuation over stationary solutions of the weighted objective. Use homogeneous objective weights so that the algorithm can cross points where the conventional ratio of task weights diverges, then store the resulting network checkpoints as an approximate Pareto set.
Replace an unconstrained simplex router or differentiable mixture layer with a resource-cost-aware router whose learned costs satisfy the paper's monotonicity curvature condition. Use a Euclidean-regularized Frank–Wolfe oracle to update routing probabilities, which should reduce cycling and sensitivity when several examples or agents compete for the same experts.