Architecture ideas

Research ideas extracted from mathematics papers, categorized as Architecture.

Unverified 2026

GQL Safe Residual Layer

Insert a scalar flux-correction-style limiter after a neural operator predicts a conservative state or residual. Interpolate between a known-admissible baseline state and the learned high-order candidate, choosing the largest coefficient that satisfies a geometric family of linear inequalities encoding positive density, positive pressure, and subluminal velocity. This retains as much of the neural prediction as possible instead of independently clipping physical variables.

Useful6/10
Difficulty5/10
Novelty7/10
Paper: GQL-Based Physical-Constraint-Preserving High-Order Finite Difference Schemes for Special Relativistic Hydrodynamics in Arbitrary Dimensions arXiv:2606.31992
Unverified 2026

Positive Mellin Mixture Gate

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.

Useful6/10
Difficulty4/10
Novelty8/10
Paper: Order-Moment Transport and Hankel Determinants in Special-Function Inequalities arXiv:2606.31647
Unverified 2026

Fast-Class MoE Router

Replace a flat stochastic MoE router by a two-level continuous-time routing model: experts within a group communicate rapidly, while transitions between groups occur slowly. Use the effective class-level stationary distribution as a soft load-balancing prior, reducing routing oscillation while preserving expert specialization.

Useful6/10
Difficulty6/10
Novelty6/10
Paper: The Invariant Measure of Multiscale Markov Chains via Fast Arborescence Factorization arXiv:2606.31596
Unverified 2026

Completely Monotone Multiscale Attention Decay

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.

Useful6/10
Difficulty4/10
Novelty7/10
Paper: Bernstein Functions at Work: Coalescents, Copulas, and Subordination arXiv:2607.04467
Unverified 2026

Markov-coded synchronization-robust latent streams

Replace an independently generated discrete latent sequence or redundancy stream with a symmetric two-state Markov source whose transition probability is tuned or learned. Train the downstream transformer to reconstruct the semantic target after random insertions, while using the paper's insertion-capacity expression to select the latent rate and redundancy budget. The representation should preserve information under timing drift, repeated tokens, and inserted distractors better than iid token…

Useful6/10
Difficulty5/10
Novelty7/10
Paper: The Insertion List-Decoding Capacity and an Improved Bound on the Deletion List-Decoding Capacity arXiv:2607.03989
Unverified 2026

Path-Holonomy Attention

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.

Useful6/10
Difficulty5/10
Novelty7/10
Paper: Noncommutative Cluster Varieties and Moduli Spaces of Local Systems arXiv:2608.27284
Unverified 2026

Reversible Matrix Cluster Layer

Construct a latent layer whose node states are small positive-definite matrices and whose local updates follow a weighted cluster exchange relation rather than an unconstrained affine map. The update is reversible when the old state is retained, while noncommuting matrix products preserve relational structure that scalar cluster variables cannot represent.

Useful6/10
Difficulty6/10
Novelty8/10
Paper: Noncommutative Cluster Varieties and Moduli Spaces of Local Systems arXiv:2608.27284
Unverified 2026

Minkowski-Additive Convex Latents

Store a convex object as a direction-indexed vertex tuple and implement composition of objects through componentwise Minkowski addition and nonnegative scaling. This creates a structured residual or compositional layer where convexification is nonexpansive, making perturbation amplification controllable and avoiding repeated generic geometric optimization.

Useful6/10
Difficulty4/10
Novelty8/10
Paper: Galerkin approximations to the space of convex bodies by polytopes in nondegenerate V-representation arXiv:2608.26615
Unverified 2026

Dual-Rail Ratio Arithmetic Layer

Replace scalar neural activations by pairs of nonnegative channels whose ratio represents the signed or unsigned activation. Implement multiplication and addition through pair algebra, and renormalize each pair because the representation is invariant under multiplying both rails by the same positive scalar. This creates an explicitly bounded, cancellation-aware arithmetic layer for deep multiplicative MLPs, rational networks, and neural fields.

Useful6/10
Difficulty5/10
Novelty7/10
Paper: Instantaneous arithmetic computation via ratio-encoding in chemical reaction networks arXiv:2608.26305
Unverified 2026

Private spectral whitening front-end

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.

Useful6/10
Difficulty5/10
Novelty7/10
Paper: On the privacy cost for dependent Gaussian data: spectral density estimation under local differential privacy arXiv:2608.24847
Unverified 2026

Noncrossing Brace Attention

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.

Useful6/10
Difficulty6/10
Novelty6/10
Paper: Braces on the cohomology of noncrossing 2-partitions arXiv:2608.24820
Unverified 2026

Exact Elasticity-Complex Message Passing

Construct a mesh neural network with node, edge, face, and cell feature spaces modeled on the four spaces of the discrete elasticity complex. Replace unconstrained cross-order message passing by fixed incidence and geometric operators whose compositions vanish exactly, so gradient-like, incompatibility-like, and divergence-like features cannot contain algebraically spurious components.

Useful6/10
Difficulty6/10
Novelty6/10
Paper: A Vector-Valued Co-Chain/Chain Complex Associated to the Elasticity Complex arXiv:2608.23829
Unverified 2026

PSD Spectral CNN Block

Parameterize a multi-channel two-dimensional convolutional operator through a learned filter bank B, then use the composed operator B*B as the layer response. Its Fourier response is positive semidefinite exactly at every spatial frequency, enabling stable smoothing or diffusion-like residual updates without frequency-grid penalty terms.

Useful6/10
Difficulty4/10
Novelty5/10
Paper: Addendum to "Factoring non-negative operator valued trigonometric polynomials in two variables" arXiv:2608.23073
Mechanism failed 2026

Critical-depth sparse attention

Represent candidate two-dimensional attention windows as dyadic rectangles and penalize local regions where many deeply embedded windows overlap. Use complementary horizontal and vertical depth exponents rather than independently penalizing one coordinate. The resulting router should reduce pathological concentration of sparse attention computation while preserving access to multiscale context.

Useful6/10
Difficulty6/10
Novelty7/10
Paper: Critical two-depth Journé packing for bi-parameter and Zygmund rectangles arXiv:2608.22628
✓✓ Beats tuned baseline 2026

Cyclic Lie-Bracket Residual Block

Replace one deterministic residual update with a short cyclic composition of learned vector fields evaluated for randomized, short run times. Because finite compositions of noncommuting flows generate directional-derivative and Lie-bracket terms, changing the cycle order gives the network an explicit, low-cost way to learn drift directions that are unavailable from the individual vector fields alone.

Useful6/10
Difficulty5/10
Novelty7/10
Paper: Diffusion limits of cyclic finite-velocity random motions along vector fields arXiv:2608.22514
Failed on benchmark 2026

Laplace-Heterogeneous MoE Routing

Replace the usual hand-designed expert-load penalty with a heterogeneous survival penalty derived from a susceptibility distribution. Each expert receives an availability factor q_e=G(A_e), where A_e is its cumulative recent routing pressure and G_e is a learned or fixed mixture of exponentials; highly used experts are suppressed smoothly, while heterogeneous experts can have different resistance to pressure. The mixture produces adaptive curvature and long-tailed penalties that may reduce…

Useful6/10
Difficulty4/10
Novelty6/10
Paper: From Individual-Based Stochastic Epidemics to Heterogeneous SIR Equations arXiv:2608.22122
Unverified 2026

Cycle-stable graph MoE routing

Use the graph-coloring stability concept to route graph nodes to experts. Nodes rank experts by router logits, adjacent nodes are constrained to use different experts, and a blocking cycle is a directed cycle in which every node prefers the expert currently assigned to the next node. Eliminate profitable feasible cycles or penalize their existence so routing reaches a locally stable assignment instead of oscillating between equally plausible expert allocations.

Useful5/10
Difficulty6/10
Novelty9/10
Paper: Graph Coloring with Color Preferences arXiv:2609.00569
Unverified 2026

Zero-mode MPS front-end for constrained sequence models

Build a constrained autoregressive model whose initial logits are generated from a translation-invariant MPS associated with a local zero-mode construction. The MPS supplies a structured valid distribution before a Transformer residual is added, so the model starts on the constraint manifold instead of learning validity through a penalty.

Useful5/10
Difficulty6/10
Novelty6/10
Paper: Efficient search for excitable zero-modes in constrained systems arXiv:2608.31165
Unverified 2026

Principal-bundle gauge-fixed Hamiltonian network

Represent a time-dependent Hamiltonian system on the reduced state $(q,t,p_q)$ rather than on the redundant extended state $(q,t,p_q,p_t)$. A neural Hamiltonian section predicts one canonical representative of each affine cotangent fiber, while an optional symmetry loss enforces consistency under transformations that translate time.

Useful5/10
Difficulty5/10
Novelty8/10
Paper: Reduction of symmetric time-dependent Hamiltonian systems I: presymplectic principal $\mathbb{R}$-bundles arXiv:2608.28278
Unverified 2026

Variation-Diminishing Channel Mixer

Constrain a channel-mixing layer to be a product of nonnegative bidiagonal matrices, rather than an unconstrained dense matrix. The resulting totally nonnegative operator is predicted not to increase sign oscillations in ordered channel features, potentially reducing high-frequency feature noise and making deep stacks more stable.

Useful5/10
Difficulty4/10
Novelty8/10
Paper: The Exact Maximum of the Spectral Sum of Graphs arXiv:2607.23081
Unverified 2026

Level-Set Balanced Sparse Mixer

Partition activations into dyadic magnitude bands and allocate sparse connectivity separately to heavy and diffuse coordinates. Protect high-magnitude coordinates with more reliable connections while using randomized flat connectivity for the many small coordinates, keeping the total number of nonzeros fixed.

Useful5/10
Difficulty6/10
Novelty7/10
Paper: Level-set entropy and sparse randomized embeddings arXiv:2607.23017
Unverified 2026

Weak-Bounded Riesz Attention

Replace one local spatial aggregation in a CNN or vision transformer with a discretized Riesz potential whose kernel is proportional to $\|x-y\|^{-(n-s)}$. Normalize the layer using the paper's sharp weak-type constant and penalize empirical violations of the resulting tail bound, encouraging nonlocal context without allowing a small set of pixels or tokens to generate arbitrarily large responses.

Useful5/10
Difficulty5/10
Novelty7/10
Paper: Sharp constants for weak estimates of Riesz Potentials when $0<s<\min\{n,2\}$ arXiv:2608.31043
Unverified 2026

Cross-Channel Vector Wavelet Front End

Replace a channelwise wavelet or strided-convolution front end with vector-valued wavelet filters that deliberately pair different scalar wavelets across channels. The resulting subbands retain compact-support multiscale structure and can be recombined exactly, while a small learned 1x1 mixing layer operates on the cross-channel coefficients instead of learning a full expensive convolution at every scale.

Useful5/10
Difficulty5/10
Novelty4/10
Paper: Vector-Valued Wavelet Bases as Hilbert $\mathbb{M}_m(\mathbb{R})$-Module Bases: A Construction from Scalar Wavelets arXiv:2608.30589
Unverified 2026

Uniformly bounded Jacobi spectral features

Replace raw powers or unconstrained polynomial spectral features with normalized Jacobi features whose amplitude is provably bounded on the entire input interval. Use trainable mixtures of these features in a positional encoding, graph spectral layer, or MLP front end, while preserving the theorem's normalization and optionally constraining the learned mixture norm.

Useful5/10
Difficulty4/10
Novelty7/10
Paper: The Koornwinder--Kostenko--Teschl Conjecture for Jacobi Polynomials and the Discrete Laguerre Phase Transition arXiv:2608.30486