Unverified
2026
Add a per-head controller that adjusts attention sharpness from the observed separation between within-cluster and cross-cluster token similarities. When a positive overlap gap becomes large, the controller lowers the head temperature to prevent exponentially localized attention and rank collapse; when the gap is small, it permits sharper attention so useful structure can form.
Useful6/10
Difficulty4/10
Novelty5/10
Unverified
2026
Use the shared-marginal KL projection to turn token-to-expert routing into a low-rank, exactly balanced assignment rather than relying only on an auxiliary load-balancing penalty. Tokens retain normalized routing distributions while the shared latent marginal enforces consistent aggregate usage across two independently learned routing factors.
Useful6/10
Difficulty6/10
Novelty6/10
Unverified
2026
Grow a mixture-of-experts layer by splitting one expert into two children while conserving its routing mass, and choose the split ratio to minimize the worst imbalance over all intermediate expert counts. Use the paper's sharp threshold as a hard design target: with n experts, some intermediate stage must have capacity ratio at least D_n = 2^{1-1/\lceil n/2\rceil}, so schedules substantially below this are impossible rather than merely difficult to discover. Initialize child router logits with…
Useful6/10
Difficulty5/10
Novelty8/10
Unverified
2026
Replace a Gaussian or point-estimate regression head with a heteroscedastic Student-t head whose scale and degrees of freedom depend on the learned state. This gives the model a principled way to absorb abrupt, nonmonotone events and operating-condition shifts without forcing the central degradation trend toward rare extreme residuals.
Useful6/10
Difficulty3/10
Novelty4/10
Unverified
2026
Replace a uniformly sampled trajectory sequence by a binary temporal partition whose intervals are split only when the observed trajectory makes an excursion larger than a threshold. Encode one summary token per retained leaf, optionally including duration and endpoint displacement, so smooth trajectory regions receive fewer tokens while rapidly changing regions retain resolution.
Useful6/10
Difficulty4/10
Novelty6/10
Unverified
2026
Replace dense token-to-token attention in selected layers with communication through a small number of multi-token hyperedges. Each hyperedge aggregates its incident token states and broadcasts the resulting message back to those tokens, allowing higher-order interactions while reducing the number of pairwise links. Reconstruct hyperedges periodically from cumulative token displacement so stable tokens retain useful groups while rapidly changing tokens are regrouped.
Useful6/10
Difficulty6/10
Novelty5/10
Unverified
2026
Replace top-k token pruning by greedy maximization of a diversity-aware monotone submodular utility under a spacing or coverage constraint. The selector repeatedly chooses the feasible token with the largest marginal utility, avoiding the redundant-token failure mode of independent score ranking while inheriting a constant-factor approximation guarantee under the stated 2-system abstraction.
Useful6/10
Difficulty5/10
Novelty6/10
Unverified
2026
Replace a conventional feature-pyramid sum by a bounded multiscale differential transform. At each scale, subtract a blockwise conditional expectation from a local average, then combine these residuals with bounded coefficients. Add a penalty on the largest interval response so that contributions from adjacent scales cannot accumulate destructively or explosively.
Useful6/10
Difficulty5/10
Novelty6/10
Unverified
2026
Replace a learned dense token-mixing matrix or residual-state transition with a sparse diffusive mixer whose Laplacian has a deliberately small largest Jordan block. Balance the two chain lengths around the central coupling/core, because the paper proves that this minimizes the worst defective transient among the tridiagonal family. Use a scalar residual step size to move the non-consensus spectrum inside the unit disk while preserving the sparse structure.
Useful6/10
Difficulty5/10
Novelty7/10
Unverified
2026
Construct an orthogonally equivariant residual map on symmetric feature matrices whose update is strongly monotone by adding the identity to a monotone isotropic tensor function. This provides a stability-controlled matrix block and a route to well-behaved inverse or fixed-point inference, rather than relying only on unconstrained residual weights.
Useful6/10
Difficulty6/10
Novelty7/10
Unverified
2026
Replace an unconstrained matrix nonlinearity on small symmetric feature blocks with the isotropic spectral lift of a permutation-equivariant monotone map on eigenvalues. The layer remains orthogonally equivariant, while the paper's equivalence transfers a scalar inner-product monotonicity certificate from eigenvalue space to the full matrix space.
Useful6/10
Difficulty5/10
Novelty6/10
Unverified
2026
Add a causal memory branch whose weights are generated by the paper's power-type Volterra kernel rather than learned independently at every lag. Learn or softly constrain the exponents so the model can select rough short-memory behavior or smoother long-memory behavior while using only a few parameters. The branch can be implemented as a truncated causal convolution, a multiresolution approximation, or a recurrent state-space realization.
Useful6/10
Difficulty6/10
Novelty6/10
Unverified
2026
Use attention-graph hitting times to identify tokens whose information has not mixed through the network, then route only those tokens through additional Transformer blocks. Tokens with fast reachability exit early, while slow or isolated tokens receive more computation.
Useful6/10
Difficulty6/10
Novelty7/10
Unverified
2026
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.
Useful6/10
Difficulty5/10
Novelty7/10
Unverified
2026
Use enumerated weighing matrices as sparse orthogonal channel-mixing operators inside MLPs or residual blocks. Their ternary entries reduce multiplication to signed additions, while exact orthogonality prevents amplification or attenuation of feature norms; a trainable fixed-support version can recover expressivity without giving up computational sparsity.
Useful6/10
Difficulty5/10
Novelty6/10
Unverified
2026
Use the paper’s observation that the fully irreducible vertex is approximately local after crossed-channel ladders are removed to build a block-local curvature correction for neural-network optimization. Estimate a cheap bare covariance and subtract the inverse full covariance to obtain a local irreducible correction, avoiding a dense four-point model while retaining interaction effects that ordinary diagonal preconditioners miss.
Useful6/10
Difficulty6/10
Novelty7/10
Unverified
2026
Use the support graph of a sparse attention or routing mask as a feasibility object before applying normalization. Reject or minimally repair masks whose neighborhoods violate Hall's condition, ensuring that every selected query/key subset has enough reachable mass to support the desired marginals.
Useful6/10
Difficulty5/10
Novelty6/10
Unverified
2026
Represent a neural-network weight tensor by rank-one terms whose mode factors are selected from shared orthonormal bases, and impose the same basis alignment across tensor flattenings. During or after training, retain the largest coefficients to obtain a structured truncation analogous to truncated SVD. This should produce better-conditioned tensorized layers than unconstrained CP factors while preserving a directly controllable accuracy/compute tradeoff.
Useful6/10
Difficulty5/10
Novelty5/10
Unverified
2026
Use a strictly upper-triangular block operator to represent interactions between ordered layers, experts, or token groups, and compute its inverse exactly with a finite Neumann series. Because nilpotency truncates the series after a known number of blocks, the module avoids an iterative solver while retaining controlled long-range interactions.
Useful6/10
Difficulty5/10
Novelty8/10
Unverified
2026
Construct a neural layer with two independently ordered transformations and train its operators to satisfy the paper's diamond equations, so that applying direction 1 then direction 2 gives the same result as direction 2 then direction 1. Unlike ordinary weight sharing, the mixed identity permits noncommuting operators whose interaction defects cancel exactly.
Useful6/10
Difficulty5/10
Novelty7/10
Unverified
2026
Use the paper's monotone allocation viewpoint to decide which parameters receive progressively more pruning, bit reduction, or update opportunities. A van der Corput order prevents a fixed early prefix of structured groups from being systematically over-pruned, while the logarithmic comb bound quantifies worst cumulative imbalance across nested groups.
Useful6/10
Difficulty4/10
Novelty6/10
Unverified
2026
Project each FFN residual update onto the tangent space of the current token residual direction before adding it to the stream. This preserves the component that changes representation direction while suppressing norm-only motion, which may reduce residual-norm drift and aggregation-induced representation collapse.
Useful6/10
Difficulty4/10
Novelty6/10
Unverified
2026
Replace a learned sequence-mixing matrix with a structured lower-triangular Sprugnoli operator whose square is exactly the identity. Applying the same operator in reverse reconstructs activations exactly, so it can be used as a reversible Transformer mixer or reversible channel permutation while retaining nontrivial long-range mixing.
Useful6/10
Difficulty6/10
Novelty7/10
Unverified
2026
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.
Useful6/10
Difficulty3/10
Novelty6/10