81–100
For finite oriented graphs, a rounded-up one-third minimum semidegree guarantees a directed cycle of length 3q through each vertex once q ≥ 2 and the graph meets an explicit size cutoff.
In matched simulations, the molecular model was stiffer and its plastic modes coupled more strongly to volume-changing deformation; a proposed correction narrowed several gaps.
Under small, constrained disturbances, a one-dimensional periodic model converges to a selected two-interface phase-transition state.
A two-step method transferred learning-rate settings across mixture-of-experts models and extrapolated them to a much longer training horizon, but the final choice was not exhaustively tested.
In a synthetic test, matched systems that saw only assigned evidence generally solved unseen combinations more accurately, but the full preregistered battery narrowly missed its target.
EchoCoT reached a 66.4% near-verbatim success rate on one accessible model, while proprietary-model tests relied mainly on reported lengths and summaries.
A mathematical analysis builds a hierarchy of constraints and finds a strict gap between distillation and cost for one explicit quantum state.
Two proposed solvers estimate a camera’s position, orientation and unknown focal length from sparse feature information, with favorable results on synthetic tests and visual-localization benchmarks.
A nested sampling framework relaxes exact polynomial quadrature, then uses canonical inversion to improve reconstruction in sphere and torus tests.
In simulations, a receiver given angle information reached five index bits; an AoD-free version saturated at three, while the baseline antenna speed was 385 km/h.
Orthogonal JEPA records higher task scores in several comparisons and lower reported errors in molecular rollouts, while the paper warns that orthogonal factors are not automatically independent or interpretable.
A real-time strategy using GOES X-ray readings became more accurate near the flare peak, while its uncertainty coverage fell for X-class events.
A constructive proof on the three-dimensional torus shows that weak solutions can approach many prescribed endpoint states, while nonuniqueness appears for a dense set of initial data.
A symbolic classification finds four exceptional graphs beyond the tangency family in a quadric-based scheme; the full union problem in higher odd dimensions remains open.
Its framework reaches a finite global construction for one frozen generation, but does not show that repeated rounds of the process terminate.
Calculations of the Dicke model recover distinct exponents in closed and open conditions, while ramp-speed results pass a Kibble–Zurek consistency check.
V-REX reached reported micro-F1 scores of 0.34 to 0.4 on proprietary veterinary data, but the benchmark did not test clinical use.
A mathematical study says preserving rankings under every allowed transformation leaves little room for non-dictatorial objective functions.
Tests on four molecular systems reproduced several thermal spectral features, while the method’s scope remains limited when tunnelling or wavepacket splitting dominates the spectrum.
Fourier and interpolation-based reconstructions often showed lower mean squared error in finite-shot tests, while hardware results remained entangled with calibration effects.