✗ Failed on benchmark
2026
Replace generic learning-rate selection in decentralized or federated gradient tracking with a low-dimensional minimax search over the exact scalar-mode pole radius. The optimizer chooses the step size that minimizes the worst predicted contraction over the observed graph spectrum and an estimated curvature interval, rather than relying only on conservative global bounds.
Useful7/10
Difficulty5/10
Novelty6/10
✗ Failed on benchmark
2026
Use computable upper and lower error bounds to reject quantized-depth configurations that cannot reach the desired accuracy before training. The planner separates irreducible library mismatch from finite-depth synthesis, codebook metadata, and execution errors, then selects the smallest depth and metadata budget whose estimated bound passes the target.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Use a learned quasipotential barrier as feedback for optimizer noise and restart control. Increase stochasticity when training is trapped in a high-loss metastable basin and reduce it near a desirable basin, with switching thresholds determined by the estimated barrier rather than by a fixed patience schedule.
Useful7/10
Difficulty6/10
Novelty7/10
✗ Failed on benchmark
2026
Replace independently restarted proximal-gradient or quasi-Newton solves for a composite neural objective with a curvature-recycling Douglas–Rachford loop. The previous proximal state, residual, and limited-memory BFGS curvature pairs are transported to the next proximal center, reducing expensive loss and gradient evaluations while retaining the cheap nonsmooth proximal operation.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Replace a single optimizer trajectory by N parameter particles and optimize the time until the first particle reaches a target loss or reward threshold. Use distinct interaction regimes: bounded normalized interactions should provide only the usual logarithmic extreme-search improvement, whereas unnormalized coherent force accumulation and stochastic pairwise kicks should produce distinct 1/N and 1/(N ln N) first-hit laws.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Replace ordinary gradient descent in a chosen approximately linear parameter block with gradient descent plus a controlled negative quadratic penalty, and stop before the unstable directions explode. The finite-time spectral filter can amplify well-supported directions while retaining shrinkage or limited exposure on weak directions, which is unavailable to a stable negative-ridge endpoint.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Use the recent history of generator outputs as a controllable training window instead of fixing the replay-memory depth globally. Estimate how quickly each fitness level improves as more same-level examples enter the window, and increase memory only when the measured escape probability improves enough to justify the extra stale data.
Useful7/10
Difficulty5/10
Novelty6/10
✗ Failed on benchmark
2026
Replace the usual parameter-plus-momentum Langevin state with a three-level chain consisting of parameters, velocity, and acceleration, while injecting Gaussian noise only into the highest auxiliary state. At a saddle, the escaping direction has a positive rate given by a cubic characteristic equation; use this rate to choose damping or adapt the temperature so that basin escape is accelerated without making the dynamics unstable.
Useful7/10
Difficulty5/10
Novelty6/10
✗ Failed on benchmark
2026
Replace a neural-network optimizer's globally fixed learning-rate geometry with an adaptive quadratic trust region. At every update, construct a local curvature model, accept or reject the step using the ratio between realized and predicted loss decrease, and expand or contract the radius accordingly; the same controller should automatically become conservative in nonconvex regions and Newton-like near a well-conditioned minimum.
Useful7/10
Difficulty6/10
Novelty6/10
✗ Failed on benchmark
2026
Construct overlapping attention windows from a block design instead of using one dense sequence-by-sequence attention matrix. Every token appears in exactly $r$ windows and every token pair co-occurs in exactly $\lambda$ windows, giving uniform coverage and avoiding the uneven connectivity of arbitrary sparse masks.
Useful7/10
Difficulty5/10
Novelty6/10
✗ Failed on benchmark
2026
Split hidden dynamics into relaxation bands when the Jacobian spectrum has a gap, evolve each band with its own timescale, and retain an explicit cross-band exchange term. This yields a principled dual-timescale RNN or SSM rather than choosing fast and slow branches heuristically.
Useful7/10
Difficulty6/10
Novelty6/10
✗ Failed on benchmark
2026
Replace a stationary optimizer by a periodic two- or multi-phase schedule, such as alternating large and small learning rates, SGD and momentum, or gradients from different loss components. Stability is assessed over the complete period using the product of phase-wise linearized update maps, allowing a phase that is individually expansive to be safely combined with a contracting phase.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Replace a capacity-penalty-only MoE router with a nonnegative shadow price for each expert, capacity bucket, or hardware resource. Route each token using predicted utility minus the relevant price, while computing a decomposed optimistic objective that certifies how much utility remains above the feasible routed value.
Useful7/10
Difficulty4/10
Novelty5/10
✗ Failed on benchmark
2026
Compress a transformer KV cache by selecting actual past tokens whose key or hidden-state columns form a stable basis for all cached tokens. Instead of retaining tokens with the largest attention scores or leverage scores independently, compute rank-revealing pivoting of the leading right-singular-vector matrix and retain its pivot columns, then evaluate attention using the representatives plus an optional low-cost residual correction.
Useful7/10
Difficulty5/10
Novelty6/10
✗ Failed on benchmark
2026
Use a two-mode optimizer: a learned preconditioned update for normal training and a bounded contractive fallback when the learned update is predicted to increase a monitored energy. Use separate entry and exit thresholds so minibatch noise does not cause rapid switching.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Replace repeated time-stepping of a stiff linear state-space block with a quadrature approximation to its inverse Laplace transform. The layer propagates a hidden state using a small set of complex shifted linear solves, which can be batched and reused across many time steps or parameter values.
Useful7/10
Difficulty6/10
Novelty6/10
✗ Failed on benchmark
2026
Replace the assumption that a minibatch gradient is fully Gaussian by a Gaussian center plus an explicit single-example big-jump correction. At each update, estimate the distribution of per-example gradient projections along the proposed update direction and use the predicted aggregate tail probability to reduce the step size or increase clipping only when the minibatch is in its non-Gaussian crossover regime.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Reparameterize a recurrent or state-space layer so that its hidden-state update contains an explicit stabilizing feedback controller, while the neural network learns only a residual control in the feedback coordinates. Choose K to reduce finite-horizon state-propagation amplification, suppressing exploding hidden states and gradients on long sequences.
Useful7/10
Difficulty5/10
Novelty5/10
✗ Failed on benchmark
2026
Partition tokens into a rectangular grid and use Latin-square labels to define several sparse attention heads. Each head attends only to tokens sharing a row, column, or Latin label, while orthogonality guarantees that every pair of labels occurs at most once, reducing systematic blind spots and repeated collisions. The resulting masks are deterministic, reusable across examples, and can be generated without learned routing scores.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Regularize the local recurrent Jacobian by its spectral radius rather than imposing the overly conservative operator-norm condition $\|J\|_2<1$. This permits useful non-normal updates with transient amplification while explicitly pushing the asymptotic dynamics toward a stable fixed point.
Useful7/10
Difficulty7/10
Novelty7/10
✗ Failed on benchmark
2026
Construct a graph of cheap prediction agents or reasoning traces and use a sparse set of expensive verifier calls as graph anchors. Select the next verifier location by the exact reduction in a trace-inverse coherence objective per unit cost, rather than by uncertainty or random sampling. This creates a budgeted mixture-of-agents architecture that can spend computation where it most improves global consensus.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Augment an optimizer with a measurable redistribution time for its internal state and compare it with the time scale of the changing gradient field. Use the resulting Damkohler number to interpolate between a fast quasi-static preconditioner and a history-preserving, non-equilibrium update, rather than applying one optimizer regime throughout training.
Useful7/10
Difficulty5/10
Novelty7/10
✗ Failed on benchmark
2026
Replace quadratic self-attention over a sequence with a bank of K auxiliary exponentially decaying states whose rates are fitted directly from the empirical autocorrelation of the sequence features. Each mode captures a distinct time scale, so the module can represent short- and long-range dependencies with O(TK) computation and O(K) recurrent memory rather than storing all previous tokens. Constrain decay rates to be positive and use the paper's extended Markovian block structure to obtain a…
Useful7/10
Difficulty5/10
Novelty4/10
✗ Failed on benchmark
2026
Replace a weight-tied residual or neural-ODE stepper with an explicit Runge–Kutta method satisfying the reused-last-stage conditions. The final derivative is evaluated at the exact endpoint and becomes the first derivative of the next step, saving one expensive neural-vector-field call per step while preserving the designed integration order.
Useful7/10
Difficulty5/10
Novelty5/10