Adam with weight decay induces a shell-core topology in parameter space whose stopping-time geometry yields grokking scaling laws for learning rate, batch size, and ℓ₂ coefficient.
Bender and Steven A
9 Pith papers cite this work, alongside 5,000 external citations. Polarity classification is still indexing.
9
Pith papers citing it
5,000
external citations · OpenAlex
citation-role summary
background 3
citation-polarity summary
years
2026 9roles
background 3polarities
background 3representative citing papers
Real-analytic perturbations of AdS4 black branes exhibit stretched-exponential decay exp(-c t^{5/6}) controlled by the large-k tail of the quasinormal mode spectrum.
High-energy string amplitudes have asymptotic expansions governed by Bernoulli numbers, upgraded via resurgence to transseries whose Stokes data encode non-perturbative monodromy between kinematic regions.
DenseAMs show tradeoffs between entropy production, retrieval accuracy, and speed at intermediate loads, with a new failure mode in higher-order networks at finite temperature.
The Random-Action-Removal algorithm solves n-state, m-action 2-player games in e^{O(√(n ln(m/n)))} time, improving the previous e^{O(√(n ln(m/√n)))} bound by exploiting the hypercube structure of the game.
Several observables in planar N=4 SYM, including the octagon anomalous dimension and Bremsstrahlung function, admit a once-subtracted dispersion representation over a positive measure in the coupling.
The minimal number of dynamical degrees of freedom in regularised scalar field theory scales with area, governed by the count of distinct normal-mode frequencies below the ultraviolet cutoff.
At large chemical potential the Gross-Neveu model enters a crystalline phase in which a-particle bound states condense, producing a periodically oscillating chiral condensate governed by two new scales Λ_n and Λ_c that replace the usual Λ.
GPU port of entropy-stable DG Euler solver with non-conservative buoyancy terms reaches nearly 70% of 64-bit peak on A100 volume kernels, delivers 10x speedup and 13x better energy efficiency versus CPU, and preserves symmetry-based flux savings.
citing papers explorer
-
A Stochastic--Geometric Theory of Scaling Laws in Grokking
Adam with weight decay induces a shell-core topology in parameter space whose stopping-time geometry yields grokking scaling laws for learning rate, batch size, and ℓ₂ coefficient.
-
After the Fluid: Subexponential Decay in AdS$_4$
Real-analytic perturbations of AdS4 black branes exhibit stretched-exponential decay exp(-c t^{5/6}) controlled by the large-k tail of the quasinormal mode spectrum.
-
Resurgence of high-energy string amplitudes
High-energy string amplitudes have asymptotic expansions governed by Bernoulli numbers, upgraded via resurgence to transseries whose Stokes data encode non-perturbative monodromy between kinematic regions.
-
Stochastic Thermodynamics of Associative Memory
DenseAMs show tradeoffs between entropy production, retrieval accuracy, and speed at intermediate loads, with a new failure mode in higher-order networks at finite temperature.
-
Improved subexponential analysis of the Random-Action-Removal algorithm for 2-player turn-based games and non-binary AUSOs
The Random-Action-Removal algorithm solves n-state, m-action 2-player games in e^{O(√(n ln(m/n)))} time, improving the previous e^{O(√(n ln(m/√n)))} bound by exploiting the hypercube structure of the game.
-
Positivity properties of observables in planar maximally supersymmetric Yang-Mills theory
Several observables in planar N=4 SYM, including the octagon anomalous dimension and Bremsstrahlung function, admit a once-subtracted dispersion representation over a positive measure in the coupling.
-
Area Scaling of Dynamical Degrees of Freedom in Regularised Scalar Field Theory
The minimal number of dynamical degrees of freedom in regularised scalar field theory scales with area, governed by the count of distinct normal-mode frequencies below the ultraviolet cutoff.
-
Perturbative, Nonperturbative and Exact Aspects of Crystalline Phases in the Gross-Neveu Model
At large chemical potential the Gross-Neveu model enters a crystalline phase in which a-particle bound states condense, producing a periodically oscillating chiral condensate governed by two new scales Λ_n and Λ_c that replace the usual Λ.
-
GPU Performance of an Entropy-Stable Discontinuous Galerkin Euler Solver with Non-Conservative Terms
GPU port of entropy-stable DG Euler solver with non-conservative buoyancy terms reaches nearly 70% of 64-bit peak on A100 volume kernels, delivers 10x speedup and 13x better energy efficiency versus CPU, and preserves symmetry-based flux savings.