Architecture ideas

Attention variants, state-space and recurrent cells, normalization and token-mixing schemes — each tested against the standard block it replaces.

Unverified 2026

Stable Magnitude Bottleneck

Insert a magnitude-only bottleneck whose output is the absolute value of a random independent-feature expansion of the latent vector. Train a decoder to reconstruct the latent representation or input modulo one global sign, while explicitly rejecting feature distributions whose normalized L1 mass is too small. The module provides a controlled way to obtain sign-invariant representations without allowing arbitrary coordinate-wise sign loss.

Useful5/10
Difficulty4/10
Novelty7/10
Paper: Stable Phase Retrieval for Spans of Independent Random Variables arXiv:2607.06693
Unverified 2026

Random-label irregular sparse attention

Construct a sparse attention mask from a fixed regular candidate graph and one scalar random label per token, retaining edge $(u,v)$ when $x_u+x_v\geq\tau$. Unlike independent random pruning, this produces correlated neighborhoods and a controllable distribution of token degrees, potentially giving some tokens broad receptive fields while retaining a fixed sparse budget.

Useful5/10
Difficulty3/10
Novelty7/10
Paper: Irregular subgraph in a regular graph arXiv:2607.06465
Unverified 2026

Patterned-Walk Graph Signature

Augment a graph neural network or graph transformer with counts of cyclic walks whose successive steps are required to be graph edges or graph non-edges according to a binary pattern. These features encode induced-subgraph structure that ordinary adjacency powers miss, and can be concatenated to the graph-level token or used as an auxiliary prediction target.

Useful5/10
Difficulty4/10
Novelty7/10
Paper: Generalized spectral closedness of $\mathcal{F}$-free graph classes arXiv:2607.06455
Unverified 2026

Peel-and-pass polynomial latent dynamics

Replace step-by-step hidden-state storage in a latent ODE, state-space model, or world model with a polynomial trajectory represented independently on short time blocks. At the end of each block, pass the next hidden state by summing temporal coefficients, allowing training and inference to discard the completed block while retaining a mathematically exact block interface.

Useful5/10
Difficulty6/10
Novelty7/10
Paper: Sparse space-time spectral methods can time-step by peel and pass arXiv:2607.06449
Unverified 2026

Cyclic Non-Backtracking Mixer

Replace a dense token or channel mixing matrix by a fixed sparse directed graph whose states are ordered pairs of symbols and whose transitions advance through a cyclic phase. Each state has exactly two allowed successors, obtained by appending a symbol different from the previous two, producing a strongly connected, vertex-transitive sparse mixer with shared local dynamics. The prescribed phase structure prevents arbitrary short-cycle routing and can act as an anti-collapse inductive bias in…

Useful5/10
Difficulty5/10
Novelty7/10
Paper: Long Directed Cycles in Vertex-Transitive Digraphs arXiv:2607.05807
Unverified 2026

Rigidity-Calibrated Set Attention

Augment pairwise attention on a set of n tokens with a rigidity operator derived from normalized pairwise directions. The operator couples infinitesimal node displacements through changes in pairwise distances, while the complete-graph theorem provides a geometry-independent eigenvalue target n/2 after spherical centering and normalization.

Useful5/10
Difficulty6/10
Novelty7/10
Paper: The Second Largest Eigenvalue of Stiffness Matrices of Normalized Complete Frameworks arXiv:2607.05472
Unverified 2026

Marginal-Preserving Difference Noise

Construct two latent variables X and Y with exactly the same marginal distribution, while forcing their difference X-Y to follow a chosen centered noise or residual law. Insert the pair into a residual, VAE, or diffusion block so that the model receives the desired perturbation without changing the marginal latent distribution at either endpoint. This creates a controlled alternative to independently sampled noise, especially when marginal drift in repeated stochastic layers is harmful.

Useful5/10
Difficulty5/10
Novelty7/10
Paper: By Law, Every Zero-Mean Risk Is the Difference of Two Equally Distributed Risks arXiv:2607.05460
Unverified 2026

Cyclotomic p-Cap Layer

Replace an ordinary token aggregation step with a p-replica cyclic-equivariant block. Features are copied into p replicas, processed by shared operators, coupled through a cap-like bilinear interaction, and projected onto cyclic invariants. An auxiliary commutation loss enforces that applying the operator before or after the p-fold lift gives similar outputs.

Useful5/10
Difficulty5/10
Novelty8/10
Paper: Noncommutative Cartier Formulae arXiv:2607.05360
Unverified 2026

Microscopic Boundary Pooling

Replace uniform set or point-cloud pooling with a microscopic weighting computed from pairwise feature-space distances. The resulting signed pooling vector should retain boundary and geometrically isolated points that ordinary mean pooling suppresses, potentially improving recognition when class information is concentrated on shape extremities or rare local configurations.

Useful5/10
Difficulty5/10
Novelty7/10
Paper: The microscopic weighting on a metric space arXiv:2607.05349
Unverified 2026

Pareto-Score Dynamic Sparse Masks

Replace a neural layer's dense parameters by a real parameter tensor and a binary activity mask, then update the mask using importance scores measured across multiple perturbation intervals. Gates that repeatedly occur in non-dominated accuracy–cost configurations receive larger scores and are more likely to be retained or activated, avoiding uniformly random sparse rewiring.

Useful5/10
Difficulty5/10
Novelty4/10
Paper: A Large-Scale Sparse Multiobjective Optimization Algorithm Based on Optimal Performance Scores arXiv:2607.04765
Unverified 2026

Noncommutative path-word pooling

Replace ordinary bag-of-events pooling for an ordered trajectory, graph walk, or token event stream with a reduced-word representation in a free group. Each event contributes a signed group word, and the model aggregates signed differences (w-1), preserving order-sensitive information while making explicitly paired local events cancel.

Useful5/10
Difficulty6/10
Novelty8/10
Paper: Homotopy index polynomials for knotoids arXiv:2607.04737
Unverified 2026

Piecewise-Symmetric Tensor Layer

Replace an unconstrained order-k weight tensor with a sum of components that are symmetric only within selected contiguous index blocks. This preserves interactions between blocks while tying parameters under within-block permutations, providing a tunable middle ground between a fully dense tensor and a fully symmetric tensor.

Useful5/10
Difficulty5/10
Novelty7/10
Paper: Piecewise Symmetric Tensors arXiv:2607.04712
Unverified 2026

Histogram-Preserving Variation Projection

Insert a rearrangement operation on scalar feature maps sampled along an ordered coordinate such as time, spatial position, or a neural-field input grid. The operation sorts values into non-increasing order, preserving the empirical histogram exactly while provably not increasing the Riesz fractional variation in the ideal one-dimensional continuous setting.

Useful5/10
Difficulty5/10
Novelty7/10
Paper: A variation on the Pólya-Segő principle in one dimension arXiv:2607.03450
Unverified 2026

Invariant Möbius latent mixer

Insert a piecewise Möbius transformation as a deterministic latent mixing layer, using the paper's exact branch structure rather than a generic unconstrained MLP. The transformation repeatedly moves points between branches while preserving a known reference density, creating a cheap chaotic mixer with analytically computable Jacobian factors. Use a truncated, normalized version in practice so that the sigma-finite invariant measure becomes a valid finite training distribution.

Useful5/10
Difficulty5/10
Novelty7/10
Paper: Dynamics of integer zeroes of homogeneous quadratic equations over $\mathbb{R}^3$ arXiv:2607.03354
Unverified 2026

Bessel Totally-Positive Attention

Replace ordinary dot-product attention logits with a strictly totally positive kernel evaluated on positive, ordered scalar coordinates attached to queries and keys. Use the modified-Bessel kernel K(x,s)=I_s(x), whose every ordered minor is positive, then row-normalize it as an attention matrix. This creates an attention operator with a mathematically enforced anti-oscillatory structure rather than merely positive entries.

Useful5/10
Difficulty6/10
Novelty8/10
Paper: Strict Total Positivity from Spectral Darboux and Toeplitz Smoothing Mechanisms arXiv:2607.02778
Unverified 2026

Matroid-circuit equivariant message passing

Represent each matroid circuit as a structured hyperedge and perform message passing from circuit embeddings back to their constituent elements. Tie all circuit-update parameters that lie in the same automorphism orbit, so relabelings preserving the matroid produce exactly relabeled hidden states rather than requiring the network to learn this symmetry from data.

Useful5/10
Difficulty5/10
Novelty6/10
Paper: Encoding matroids into quantum states arXiv:2607.02736
Unverified 2026

Smooth-Plus-Boundary-Lifting Network

Represent the prediction as a sum of a smooth interior branch and a fractional boundary branch: u_theta(x)=u_int_theta(x)+d(x)^a u_bd_theta(x). This mirrors the paper's direct-sum solution structure and allocates separate network capacity to the globally regular component and the boundary layer.

Useful5/10
Difficulty5/10
Novelty8/10
Paper: The structure of solution spaces for fractional-order operators, with gradient estimates arXiv:2607.02312
Unverified 2026

Tree-Cone Distribution Head

Represent a neural network's categorical output over a rooted tree using cumulative probability mass on each rooted subtree. Train pairs of examples with a stochastic-dominance loss that compares these subtree masses, avoiding enumeration of all upper sets and making hierarchical monotonicity explicit. This is suitable for taxonomies, severity levels, hierarchical intents, and structured world-model states.

Useful5/10
Difficulty3/10
Novelty4/10
Paper: Characterizing finite posets whose probabilistic powerdomain are RB-domains arXiv:2607.02231
Unverified 2026

Subcritical Preferential Sparse Attention

Construct a sparse attention or message-passing mask by sampling edges with preferential weights (d_u+alpha)(d_v+alpha), while keeping the edge count below the predicted connectivity threshold. This creates hub-like local communication patterns but prevents one giant component from forcing dense information mixing, reducing attention cost and potentially mitigating oversmoothing.

Useful5/10
Difficulty5/10
Novelty7/10
Paper: Sharp Asymptotics for the Largest Component in the Subcritical Regime of Preferential Attachment Without Vertex Growth arXiv:2607.00731
Unverified 2026

Rank-safe Bernoulli layer initialization

Use the Bernoulli corank asymptotic to choose sparsity for binary or sparse linear layers and reject initial matrices with excessive numerical rank deficiency. The layer should also explicitly prevent zero columns, because the paper's probability law indicates that zero-column events are a leading mechanism behind large corank in the sparse regime.

Useful5/10
Difficulty4/10
Novelty5/10
Paper: Rank deficiency of Bernoulli random matrices for growing corank arXiv:2607.00495
Unverified 2026

SURE-Adaptive Derivative Front End

Prepend an adaptive Savitzky-Golay derivative bank to a temporal neural network. For each input channel and derivative order, select the local window by minimizing Stein's unbiased risk estimate, then concatenate the raw signal with the estimated derivatives. This supplies denoised velocity and acceleration features without requiring clean derivative targets or forcing the backbone to learn unstable finite-difference filters.

Useful5/10
Difficulty3/10
Novelty6/10
Paper: PDE Identification Using Noise Adaptive Differentiation in Strong Form (S-IDENT) arXiv:2606.31776
Unverified 2026

GKP Log-Concave Lag Mixer

Generate temporal attention or convolution weights with the Graham–Knuth–Patashnik recurrence instead of learning every lag weight independently. For nonnegative recurrence parameters, the resulting lag sequence is strongly log-concave, so its normalized kernel is naturally unimodal and suppresses high-frequency sign-free oscillations without requiring a separate smoothness penalty. The six parameters can be learned per head, channel group, or layer, giving O(1) learned parameters for an…

Useful5/10
Difficulty3/10
Novelty6/10
Paper: Log-concavity and log-convexity in the theory of the Graham--Knuth--Patashnik recurrences arXiv:2607.04217
Unverified 2026

PED finite-state graph layer

Add a finite-state message-passing layer that tracks local configurations corresponding to perfect edge domination or dominating induced matchings instead of transmitting unconstrained node embeddings alone. On graphs with a tree, series-parallel, or small-separator decomposition, the layer produces an exact or differentiable partition function over globally valid edge configurations, which can be used as node features, an auxiliary loss, or a structural prior.

Useful5/10
Difficulty6/10
Novelty7/10
Paper: Counting perfect edge dominating sets: extremal results and linear-time algorithms arXiv:2607.03894
Unverified 2026

Simplex-Complexity Depth Diagnostic

Estimate the simplex-based ratio of a target or learned convex piecewise-linear polytope and use the theorem \(\rho_\Delta(P)\le 2^d-1\) to choose a minimum useful ReLU depth. During training, monitor whether the learned polytope is approaching a high-\(\rho\) target; if it is, widen the model without increasing depth only when the diagnostic indicates that depth is the bottleneck.

Useful5/10
Difficulty5/10
Novelty9/10
Paper: A simplex-based measure of symmetry arXiv:2607.03815