Unverified
2026
Replace the usual independently normalized MoE router scores with a small system of tropical inequalities controlling expert load, score margins, and capacity slack. Each inequality induces a local max-plus policy selecting its currently dominant expert or constraint; policy improvement detects positive-weight cycles that would cause oscillatory routing and applies the smallest bias correction that removes them. This provides a non-differentiable but cheap controller around the router…
Useful6/10
Difficulty6/10
Novelty6/10
Unverified
2026
Use the paper's mode decomposition to choose between ordinary DIGing and augmented ATC-DIGing/AugDGM according to the communication spectrum and curvature. The augmented scheme changes the disagreement feedback gain mode by mode, which can reduce the dominant pole radius on well-connected graphs without changing the neural-network architecture.
Useful6/10
Difficulty5/10
Novelty5/10
Unverified
2026
Model an RNN as a linear state-space system in feedback with its slope-restricted activation, then search for a finite-horizon IQC multiplier instead of relying only on a spectral-radius or OZF-style condition. Penalize or reject parameter settings for which the strict IQC/LMI certificate has insufficient margin, yielding a directly testable stability criterion for long unrolled sequences.
Useful6/10
Difficulty7/10
Novelty6/10
Unverified
2026
Model a checkpointed validation metric as a finite asymptotic expansion in known decay features, such as powers of training step, and estimate its limiting value using sliding least squares. Use a ridge-stabilized fit and require agreement across multiple windows before stopping, preventing the controller from reacting to transient non-asymptotic behavior.
Useful6/10
Difficulty4/10
Novelty7/10
Unverified
2026
Replace magnitude-only pivot selection in an approximate symmetric eigensolver with a perturbation score that divides squared off-diagonal coupling by the spectral gap between the associated diagonal entries. In covariance whitening or second-order preconditioning, this should spend a limited number of rotations resolving nearly degenerate eigenspaces while ignoring harmless couplings between well-separated modes.
Useful6/10
Difficulty5/10
Novelty6/10
Unverified
2026
Replace the constant policy natural-gradient step size in an episodic RL algorithm with the paper's geometrically increasing schedule, eta_t = eta_0 (H/(H-1))^t for horizon H >= 2. The schedule is intended to offset horizon-dependent decay of policy-improvement signals and produce geometric rather than sublinear convergence in finite-horizon policy optimization.
Useful6/10
Difficulty5/10
Novelty4/10
Unverified
2026
Add an acceptance gate around transported quasi-Newton steps: use recycled curvature only when it decreases the smooth proximal merit and reduces the new residual. Otherwise discard the candidate and execute a bounded number of conservative gradient steps, making curvature reuse robust to minibatch changes and stale models.
Useful6/10
Difficulty4/10
Novelty7/10
Unverified
2026
Train a population controller as a convex mixture of neural trajectory policies, using a Frank-Wolfe step to add a new policy that minimizes the current population-cost linearization. The resulting mixture operates as a structured policy ensemble and can retain feasibility when each oracle policy satisfies the same support, action, and obstacle constraints. This is a principled alternative to directly optimizing one highly nonconvex multi-agent policy.
Useful6/10
Difficulty6/10
Novelty7/10
Unverified
2026
Add a controlled periodic phase to an optimizer, then use a near-identity normal-form transform to remove rapidly oscillating gradient components instead of allowing them to perturb parameters directly. The optimizer follows averaged drift for non-resonant frequencies but explicitly preserves Fourier components near resonance, where they can create a secular update.
Useful6/10
Difficulty6/10
Novelty8/10
Unverified
2026
Train several replicas of a neural model whose effective parameters include auxiliary coordinates, with a quadratic penalty controlling how far the replica leaves the physical parameter subspace. Low-penalty replicas can use the extra directions to bypass sharp optimization barriers, while high-penalty replicas remain close to the ordinary model; periodically exchange parameters between replicas using a replica-exchange acceptance rule.
Useful6/10
Difficulty6/10
Novelty7/10
Unverified
2026
Build an RNN from fast nonlinear units coupled through a spectrally contractive slow state. The fast component can generate rich transients, while the slow component has a provable absorbing radius because its linear recurrence contracts and its neural forcing is bounded. Cross-coupling strength is swept to detect the onset of expressive high-dimensional attractors without permitting state explosion.
Useful6/10
Difficulty5/10
Novelty6/10
Unverified
2026
Add a deterministic torus phase to a recurrent or state-space model and average predictions over a quasi-periodic phase orbit using a frequency-aware normalized window instead of a uniform average. The window is chosen to attenuate Fourier modes near the orbit frequencies, transferring the paper's cancellation mechanism to reduce coherent long-horizon oscillation and bias without requiring a highly smooth predictor.
Useful6/10
Difficulty5/10
Novelty7/10
Unverified
2026
Represent the optimizer state or recurrent hidden state as an iterated map and estimate its natural invariant measure from a sliding-window occupation histogram or feature embedding. Use convergence of long-run observable averages and distances between successive empirical measures to detect whether training has entered a stable, periodic, or chaotic statistical regime, and optionally control the learning rate without forcing pointwise convergence.
Useful6/10
Difficulty4/10
Novelty8/10
Unverified
2026
Add a slow meta-controller that governs an explicit neural-network reference, such as task weights, target-risk tradeoffs, exploration level, or an auxiliary-loss coefficient, while a fast optimizer trains the model under the current reference. The controller changes the reference only after delayed outcome evidence indicates mismatch, and should be disabled or accelerated when the evidence delay exceeds the environment's objective-drift timescale.
Useful6/10
Difficulty5/10
Novelty7/10
Unverified
2026
Use a symmetric positive-definite, non-diagonal mobility matrix to couple updates of parameter groups, analogous to drag-modified Onsager mobility coupling ionic species. Estimate local block curvature and select the learning rate from the generalized spectrum of mobility times curvature, targeting rapid loss decay without the instability of aggressively scaled diagonal optimizers.
Useful6/10
Difficulty5/10
Novelty6/10
Unverified
2026
Use BB1 for inexpensive curvature adaptation, but monitor the projective gradient state for the periodic behavior identified in the paper. When the normalized gradient and scalar step size approximately repeat after seven iterations, temporarily switch to BB2 or a damped gradient step to destroy the attracting cycle, then return to BB1.
Useful6/10
Difficulty5/10
Novelty7/10
Unverified
2026
Replace a fixed learning-rate schedule by a finite-horizon feedback controller whose action depends on a noisy estimate of the current optimization state and its uncertainty. The controller takes larger corrective steps when uncertainty is informative, but increasingly enforces the endpoint as the horizon closes, while charging an explicit cost for every intervention.
Useful6/10
Difficulty6/10
Novelty7/10
Unverified
2026
Construct a sparse routing or graph-neural architecture whose activation gates satisfy a hard-core constraint: neighboring sites, experts, or token groups cannot be active simultaneously. Compare the same local routing rule on bipartite and random regular interaction graphs; the graph structure should change the maximum usable activation dimension and may also change optimization stability.
Useful6/10
Difficulty5/10
Novelty7/10
Unverified
2026
Model each expert as a cell with occupancy q_i, vacancy n_i = 1 − q_i, and a nonzero localization floor Δ_i. Add a free-volume potential whose derivative becomes strong when an expert is poorly utilized, but remains finite because of Δ_i. Unlike ordinary entropy balancing, this mechanism predicts a quantitative inverse-vacancy regime and a measurable crossover to saturation.
Useful6/10
Difficulty4/10
Novelty5/10
Unverified
2026
Partition network parameters or output-head parameters into two coupled groups, C and D, and use a Schur-complement preconditioner rather than one shared scalar learning rate. The update approximately accounts for the response of group C before applying the curvature seen by group D, reducing the effect of cross-group gradient coupling and large condition numbers.
Useful6/10
Difficulty6/10
Novelty5/10
Unverified
2026
Replace ordinary graph propagation, which repeatedly revisits the edge it just traversed, with a directed-edge non-backtracking operator. Normalize its learned gain using an estimate of the Hashimoto spectral radius so that feature magnitudes neither explode on high-growth graphs nor vanish on sparse graphs.
Useful6/10
Difficulty5/10
Novelty6/10
Unverified
2026
Replace occasional gradient updates in a neural-network optimizer with a Gaussian-smoothed cubic-regularization step estimated entirely from loss evaluations. The cubic term stabilizes steps under uncertain curvature, while the Goldstein neighborhood interpretation makes the method robust to discontinuous or rapidly varying Hessians and gives it an explicit negative-curvature escape mechanism.
Useful6/10
Difficulty6/10
Novelty6/10
Unverified
2026
Augment each recurrent channel, feature group, or state-space stream with a latent phase oscillator and allow cross-stream coupling only when the receiving oscillator lies inside a learned or fixed phase window. The window suppresses destructive mixing outside the relevant dynamical regime while retaining Kuramoto-style attraction during the active interval, potentially improving long-horizon coherence without forcing all hidden states to synchronize continuously.
Useful6/10
Difficulty5/10
Novelty7/10
Unverified
2026
Replace independent binary early-exit or token-pruning decisions with a monotone randomized survival process for each token or expert route. A token can lose survival mass at each layer but cannot become active again; the model is trained with a reflected obstacle-style penalty that activates when the predicted value of continuing computation is below the value of stopping plus the compute cost. Mean-field statistics are computed over currently surviving tokens, making routing less sensitive to…
Useful6/10
Difficulty5/10
Novelty5/10