△ Mechanism confirmed, baseline not beaten
2026
Train a fixed-rank neural weight update Y=USV^T with a projector-splitting Runge–Kutta step instead of independently applying Adam or gradient descent to U, S, and V. The update evolves the full low-rank matrix using the neural gradient but performs QR-based factor updates, avoiding S^{-1} and remaining stable when adapter singular values collapse or cross zero. Use a common-base midpoint construction so every internal stage starts from the same U,V basis and remains rank r.
Useful7/10
Difficulty5/10
Novelty6/10
✗ Failed on benchmark
2026
Train a neural value or latent-dynamics model with temporal-difference targets before enforcing a stiff differential-equation residual, and ramp the physics weight only after the critic has become predictive. For a stochastic dynamical model, the residual is computed using the infinitesimal generator, while terminal, safe, and failure boundary conditions are imposed through separate penalties.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Mechanism failed
2026
Train a neural controller or latent dynamics model together with a finite abstraction whose cells and successor relations are optimized using a smooth reverse-simulation surrogate. Penalizing concrete-to-abstract mismatch should suppress locally inconsistent or overly expansive latent transitions, while a separate reachability containment check preserves soundness. This creates a verification-aware training signal that targets spurious branching rather than only one-step prediction error.
Useful7/10
Difficulty6/10
Novelty8/10
✗ Mechanism failed
2026
Use the paper's self-normalized martingale bound to monitor cumulative stochastic gradient noise in covariance-whitened coordinates. Convert its time-uniform confidence boundary into a trust-region multiplier: retain the normal optimizer update while the observed noise is within the boundary, and shrink or clip the update after an exceedance.
Useful7/10
Difficulty5/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Use a projected dual variable as a feedback controller for terminal feasibility rather than selecting a fixed penalty coefficient. The multiplier increases after infeasible batches and decreases after feasible batches, with an explicit cap and drift-balance diagnostic that detects whether the policy-dual loop is stable.
Useful7/10
Difficulty3/10
Novelty5/10
✓✓ Beats tuned baseline
2026
Replace unconstrained residual blocks by a nonautonomous linear backbone plus a learned nonlinear perturbation, and constrain the perturbation gain using the Green operator of the backbone. The resulting network can contain both contracting and expanding channels, but the accumulated response of the perturbation remains bounded when its Green margin is below one. A differentiable or periodically updated estimate of this margin becomes both an architecture constraint and a training monitor.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Mechanism failed
2026
Replace fixed-parameter unrolled Douglas–Rachford iterations in a differentiable convex optimization layer with a causal controller that adapts relaxation and objective-drive strength from the current residuals. The controller should accelerate early progress while enforcing admissible parameter ranges, so every individual block remains a stable relaxed splitting map rather than an unconstrained learned optimizer.
Useful7/10
Difficulty5/10
Novelty6/10
✗ Failed on benchmark
2026
Make directed edge weights trainable while constraining optimization to remain away from eigenvalue collisions of the graph Laplacian. The network can learn task-specific interaction strengths while preserving a measurable diagonalizability margin and avoiding ill-conditioned modal dynamics.
Useful7/10
Difficulty7/10
Novelty8/10
✗ Failed on benchmark
2026
Add a bifurcation-aware monitor or regularizer to a continuous-time recurrent model by evaluating the trace and determinant of its local state Jacobian along the Jacobian kernel direction. Near a nilpotent rank-one equilibrium, these quantities estimate the Bogdanov-Takens coefficients a and b, allowing training to avoid uncontrolled higher-order degeneracies or deliberately target a controlled phase transition in latent dynamics.
Useful7/10
Difficulty5/10
Novelty8/10
✗ Failed on benchmark
2026
Use the learned variational functional's second functional derivative as a consistency mechanism: equilibrium susceptibility, forces, and phase stability must all be computed from the same Hessian rather than from independently trained predictors. Penalize negative or excessively ill-conditioned Hessian modes during training, while retaining soft negative modes as a detectable phase-transition signal.
Useful7/10
Difficulty6/10
Novelty7/10
✗ Mechanism failed
2026
Train an input-generation policy or differentiable signal parameterization to produce trajectories that cover the joint input-state feature space while remaining informative for every plausible neural world model. Replace single-model experiment design by an expectation over an ensemble of models, and optimize this objective with stochastic model and trajectory samples.
Useful7/10
Difficulty6/10
Novelty7/10
✗ Failed on benchmark
2026
Replace an unconstrained recurrent or neural-ODE hidden state with a positive state driven by reaction-like polynomial flows whose rate vector is modulated by inputs or context. Train the module together with an ISS penalty so bounded gate perturbations produce a bounded hidden-state deviation, preventing long-horizon amplification while retaining nonlinear computation.
Useful7/10
Difficulty5/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Treat a neural-network training run as a time-dependent dynamical system and define scalar late-time features that distinguish convergent, oscillatory, noisy, and divergent regimes. Instead of exhaustively sweeping a two-dimensional hyperparameter grid, continue the threshold curve of a feature in the learning-rate/weight-decay or learning-rate/noise plane using a secant predictor and one-dimensional correction sweep. This produces an automatically updated stability map and can be used to keep…
Useful7/10
Difficulty4/10
Novelty7/10
△ Mechanism confirmed, baseline not beaten
2026
Use the theta-SRG of each residual-block Jacobian to regularize its gain and phase spread, rather than constraining only its spectral norm. For an implicit or deeply unrolled residual network, maintain a positive distance between the SRG enclosure of the block composition and the critical feedback point -1, giving a directly testable invertibility margin for long-horizon propagation.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Replace hand-tuned exponentially separated coefficients for multiple neural objectives with weights obtained from a local KKT certificate. For L1 hinge penalties, solve a small linear program that maximizes the smallest tier weight while enforcing approximate stationarity of the weighted objective at the current priority solution. This should preserve high-priority behavior more reliably than fixed loss weights while avoiding unnecessarily large coefficients.
Useful7/10
Difficulty6/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Replace the usual best-sample or uniform group baseline in sampled-policy training with a leave-one-out baseline weighted toward structurally dissimilar solutions. Diverse peers contribute more independent information, while near-duplicate trajectories contribute less redundant signal.
Useful7/10
Difficulty4/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Replace an unconstrained learned dynamics model in model-based reinforcement learning or neural optimal control with a Koopman-style observable lift and an explicitly estimated infinitesimal generator. Train a value network against an HJB residual formed from this generator, so the critic is constrained by the observed vector field and control directions rather than relying only on temporal-difference targets.
Useful7/10
Difficulty5/10
Novelty6/10
△ Mechanism confirmed, baseline not beaten
2026
Replace a static or uniformly random PINN collocation distribution with points generated by rolling out the model's own local feedback dynamics. For a learned scalar field V_theta(x,t), compute a control and adversarial direction from grad_x V_theta, integrate the physical dynamics forward, add controlled Gaussian exploration, and train on the resulting points together with a small uniform reservoir. This should concentrate samples near reachable boundaries, large-residual regions, and…
Useful7/10
Difficulty5/10
Novelty6/10
✗ Mechanism failed
2026
Use the interpolation SDP to synthesize coefficients for a short-memory first-order minimax optimizer with a certified worst-case contraction rate. The resulting recurrence can combine current and previous iterates and gradients, providing an offline-designed alternative to hand-tuned simultaneous descent-ascent, extragradient, or optimistic-gradient updates.
Useful7/10
Difficulty6/10
Novelty7/10
✗ Failed on benchmark
2026
Decompose a periodic recurrent or state-space model into group-symmetry sectors and temporal Fourier modes, then monitor the restricted characteristic spectrum instead of only the full Jacobian. Use the first sector whose characteristic value approaches zero or whose winding number changes to reduce the learning rate, increase damping, or deliberately activate a new dynamical mode.
Useful7/10
Difficulty6/10
Novelty7/10
✗ Mechanism failed
2026
Use the robust safety interval width as a training signal and activate conservative control before the neural policy reaches an infeasible state. The network is trained to preserve a positive reserve between competing constraints, reducing abrupt projection corrections and making the closed loop less sensitive to model and disturbance errors.
Useful7/10
Difficulty4/10
Novelty6/10
✗ Failed on benchmark
2026
Represent training near a switching condition as two locally smooth optimizer modes, such as low- and high-momentum updates or two preconditioners, with a delayed gate. Estimate the leading return-map coefficient and use the paper's scaling law to cap the delay or hysteresis width before an attracting optimization oscillation becomes large. The controller can also intentionally permit a small predicted cycle near saddles or plateaus, then remove the delay as soon as the measured cycle amplitude…
Useful7/10
Difficulty6/10
Novelty8/10
△ Mechanism confirmed, baseline not beaten
2026
Replace independent token-to-expert softmax routing with a fixed-budget congestion game. Each token group distributes a fixed routing mass across experts, while the marginal value of an expert decreases as other groups send mass there. Iteratively route toward the highest current marginal utility and exploit sorted-prefix supports to produce sparse, capacity-aware assignments.
Useful7/10
Difficulty5/10
Novelty5/10
✗ Failed on benchmark
2026
Replace a monolithic recurrent transition with multiple recurrent modules coupled through a trainable directed matrix whose spectrum is explicitly shaped for the delay-dependent master-stability region. Use heterogeneous indegrees and nonreciprocal edge weights rather than forcing symmetric or all-to-all coupling, because delays can make these structures more stable than homogeneous reciprocal coupling.
Useful7/10
Difficulty6/10
Novelty7/10