✗ Mechanism failed
2026
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.
Useful8/10
Difficulty5/10
Novelty8/10
✓✓ Beats tuned baseline
2026
Replace a purely nonlinear recurrent transition with a learned observable map followed by an explicitly linear latent evolution model. Include the original latent state and a small set of nonlinear observables, and update the linear transition online with forgetting-factor recursive least squares when the environment or task dynamics change.
Useful8/10
Difficulty6/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Build a continuous-time or discretized recurrent network whose interaction graph has trainable magnitudes and phase delays, then regularize the spectrum of the phase-corrected interaction matrix around each desired latent phase-locked state. The cosine-weighted composite matrix determines whether perturbations contract or grow, providing a computable stability margin instead of relying only on empirical exploding-gradient detection.
Useful8/10
Difficulty5/10
Novelty7/10
✗ Mechanism failed
2026
Train a latent state-space neural network so that its effective pole geometry remains consistent when identified by low-frequency moments and finite-window trajectories. Penalize disagreement between the two reductions, and penalize proximity to the oscillatory/non-oscillatory boundary, to reduce spurious ringing after distillation or context truncation.
Useful8/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Replace selected residual, recurrent, or state-space blocks by modules whose input-output Jacobians satisfy an IODP inequality throughout a prescribed activation domain. The constraint controls incremental amplification between two trajectories without requiring either trajectory to remain near one fixed equilibrium, so it should improve robustness to changing contexts and prevent exploding long-horizon sensitivities.
Useful8/10
Difficulty6/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Replace an unconstrained recurrent or neural-ODE hidden-state evolution with a parameter-conditioned vector field whose Jacobian is contractive in a learned positive-definite metric. A Lyapunov residual is added during training using the current context, time, or operating-condition vector, allowing one model to remain stable across changing regimes rather than only near one nominal point.
Useful8/10
Difficulty6/10
Novelty6/10
✗ Failed on benchmark
2026
Replace derivative-based latent-dynamics fitting with an integral regression and maintain a history stack selected by the smallest eigenvalue of its information matrix. The model should perform aggressive parameter updates only when the estimated latent regressors are sufficiently exciting, while a perturbation bound prevents false excitation caused by inaccurate hidden-state estimates.
Useful8/10
Difficulty5/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Replace an unconstrained neural ODE vector field with a learned port-Hamiltonian vector field whose energy gradient drives the dynamics, whose interconnection matrix is skew-symmetric, and whose dissipation matrix is positive semidefinite. The resulting model remains expressive through state-dependent neural matrices while guaranteeing non-increasing learned energy in the unforced case.
Useful8/10
Difficulty5/10
Novelty6/10
✗ Failed on benchmark
2026
Add an observability objective to an RNN so that a finite trajectory of selected hidden coordinates preserves information about the initial hidden state. The regularizer maximizes the smallest singular value or log determinant of the finite-horizon observation Jacobian, counteracting ReLU activation masks that erase hidden-state directions.
Useful8/10
Difficulty6/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Equip a stochastic neural ODE or recurrent state-space model with a step-size controller that explicitly checks whether the discrete-time Lyapunov exponent has the same sign as the continuous-time exponent estimate. If discretization changes an attracting mode into an expanding one, reduce the step size or use a higher-order or semi-implicit update rather than trusting ordinary Euler integration.
Useful8/10
Difficulty5/10
Novelty7/10
✓✓ Beats tuned baseline
2026
Use the paper's below-threshold bistability mechanism to distinguish local stability from actual recovery: a recurrent network may have a locally stable nominal state while a second stable state still captures trajectories. Add a perturbation-based basin test and retain stronger damping or reset actions until the network demonstrably returns to the desired branch, rather than disabling intervention immediately when the spectral threshold is restored.
Useful8/10
Difficulty6/10
Novelty6/10
✗ Mechanism failed
2026
Represent a small neural state-update map or optimizer update by polynomial constraints and certify decrease of a polynomial Lyapunov function on the nonnegative activation or state region using successive Parrilo SOS levels. Use the monotone shift-threshold construction to distinguish genuine instability from failure of a weak certificate, and raise the SOS level only when necessary.
Useful8/10
Difficulty7/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Replace the dense hidden-state trajectory of a continuous-depth or recurrent neural block by a rank-r factorization F(t) = X(t) S(t) V(t)^T, and evolve the factors with a reversible projector-splitting integrator. During backpropagation, reconstruct earlier hidden states by reversing the factor updates rather than storing all activations.
Useful8/10
Difficulty7/10
Novelty6/10
✗ Failed on benchmark
2026
At a learned switching hyperplane, replace ambiguous hard routing by a convexified vector field whose normal component is zero whenever neighboring vector fields point toward the surface. This gives a non-chattering approximation of Filippov sliding and can improve long-horizon integration near friction thresholds, impacts, and climate regime boundaries.
Useful8/10
Difficulty6/10
Novelty8/10
△ Mechanism confirmed, baseline not beaten
2026
Replace an unconstrained graph-neural latent ODE or recurrent transition with two edge-cochain states whose linear drift is Hodge-Laplacian dissipation and whose quadratic coupling is generated by a skew-symmetric anticommutator. The coupling remains expressive while cancelling from the total energy, so the long-time envelope is determined by the Hodge spectral gap rather than uncontrolled nonlinear growth.
Useful8/10
Difficulty6/10
Novelty6/10
✓✓ Beats tuned baseline
2026
Replace pointwise prediction of the next field with prediction of a learned flux followed by a discrete divergence. Combine Fourier spatial mixing with a causal temporal kernel over the recent resolved-history slab, so the model learns finite-memory closure effects while preserving local conservation exactly under periodic or compatible boundary conditions. The architecture should reduce spurious mass drift and improve autoregressive rollout stability on coarse-grained PDE data.
Useful8/10
Difficulty5/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Turn an iterative optimization or equilibrium computation inside a neural network into a differentiable layer whose backward pass solves the implicit adjoint system with conjugate gradients or GMRES using only automatic-differentiation matrix-vector products. This avoids storing unrolled iterations and avoids explicit Hessian or Jacobian construction, enabling longer solver horizons and lower-memory implicit architectures.
Useful8/10
Difficulty6/10
Novelty5/10
△ Mechanism confirmed, baseline not beaten
2026
Replace the fixed-strength measurement correction in a recurrent neural state-space model with a locally normalized correction whose amplitude is inversely proportional to the operator norm of the learned measurement Jacobian. This prevents highly sensitive learned representations from amplifying latent-state errors and should make long-horizon filtering and rollout behavior substantially less dependent on representation scale.
Useful8/10
Difficulty5/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Replace arbitrary directed-edge weights in a graph neural ODE or recurrent message-passing layer by weights constructed to make the directed Laplacian diagonalizable. This removes Jordan-block coupling, allowing the linearized graph dynamics to be represented as independent eigenmodes rather than modes with polynomial transients such as t^k exp(lambda t).
Useful8/10
Difficulty6/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Add a rare-probe channel to a recurrent or state-space model and measure its local growth around every attractor reached by the same parameters. Penalize the worst attractor-conditioned growth rate, rather than checking stability only along one training trajectory, so a model cannot appear stable in one regime while exhibiting exploding perturbations in another. The method is especially appropriate for long-horizon RNNs, neural ODEs, and autonomous world models with recurrent hidden dynamics.
Useful8/10
Difficulty6/10
Novelty6/10
✗ Failed on benchmark
2026
Replace uniformly spaced history taps in a neural state-space encoder with a fixed or learned set of non-uniform delays. Regularize the resulting delay-observation matrix to have a large smallest singular value, which makes latent-state reconstruction less sensitive to irregular timestamps and observation noise. This is directly applicable to event-based data, missing timestamps, and systems with multiple time scales.
Useful8/10
Difficulty6/10
Novelty7/10
✓✓ Beats tuned baseline
2026
Build a latent continuous-time neural model with dynamics \(\dot{z}=Az+f_\phi(z)\), where \(f_\phi\) is known, separately computed, or frozen, and \(A\) is learned exclusively from the derivative residual after subtracting \(f_\phi(z)\). Parameterize \(A\) with a truncated SVD or low-rank factorization so its eigenvalues directly predict local stability and long-horizon growth.
Useful8/10
Difficulty5/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Train and select neural ODE architectures using parameter sensitivities and Fisher information, so that a model is penalized or rejected when different parameters produce nearly indistinguishable trajectory effects. The neural component remains inside the ODE vector field, but its width, depth, and parameterization are selected using predictive error together with the smallest Fisher-information eigenvalue, effective rank, and confidence intervals.
Useful8/10
Difficulty6/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Replace an unconstrained recurrent generator with a skew-adjoint block lift whose projected first channel implements a non-skew effective generator. The hidden state evolves unitarily in the enlarged space, preventing exponential norm blow-up, while the projection can express transient amplification, damping, and non-normal dynamics unavailable to a purely orthogonal recurrent matrix.
Useful8/10
Difficulty6/10
Novelty6/10