Preprint maps when a temporal graph problem is tractable—and when it is hard
Theoretical results give efficient algorithms in restricted two- and three-rank settings, while identifying NP-complete or NP-hard cases at three and four ranks.
101–120
Theoretical results give efficient algorithms in restricted two- and three-rank settings, while identifying NP-complete or NP-hard cases at three and four ranks.
A mathematical theorem says locally integrable forcing data can correspond to stronger local integrability of a weak solution’s gradient under specified assumptions.
A fixed-background calculation finds that matching two interactions at an outer radius still leaves their different radial behavior visible in a local spinor response.
For forward-global, uniformly H1-bounded focusing Schrödinger solutions in five or more dimensions, the residual non-scattering component is confined to arbitrarily narrow spacetime cones around a finite set of asymptotic velocities.
A review of four control strategies says exploration is essential when the parameters that matter most cannot be identified without probing.
A theorem-level analysis links complete positivity with complete-isometric range structure in selected noncommutative Lp-spaces, while counterexamples mark the limits.
An analytic cosmology study reports branch-dependent stability, a lower cutoff in one case and a scalar that freezes on large scales but remains dynamic on small ones.
A scenario model found large increases in a risk indicator, but its projected scale and geography changed sharply with the climate model and emissions pathway.
A mathematical study of regular black holes and wormhole-like geometries reports strong dependence on the model’s bounce parameter in some quasinormal modes, but only small optical differences among black-hole cases.
A deployed testbed reported asymptotic rates of 0.42 and 0.93 Mbit/s, while finite-size composable security remained negative under the experimental parameters.
A backstepping-guided SAC controller performed best among learning-based methods when the simulated canal began far from equilibrium, although analytical backstepping failed in that test.
A qualitative review of 38 course syllabi and interviews with 10 instructors produced a 14-question assessment, but it has not yet been validated with students.
The system was tested on four datasets, with reported gains on complex and timestamp-focused results.
In recipe-task trials, an evidence-first planner was associated with higher completion on underspecified instructions and higher halt rates when a carrot was absent.
Calculations and simulations suggest that stress, elasticity and mechanical impedance can sort growth into different regimes, but the study contains no measurements from living tissue.
In modeled communities, a slow “soft mode” produced strong correlations only when its fluctuations stood out from the rest of the system.
The method is reported as sound and complete for a restricted query class; experiments on DBPedia and MMM exposed search bottlenecks and repeated 20-second limits.
The engineering benchmark reports lower proof-generation times and roughly one-second tallies, but it does not test whether AI answers are factually correct or useful in real-world deployment.
A mathematical analysis of spherical integral equations finds that quadrature resolution can determine whether Sloan iteration’s ideal convergence gains survive discretization.
Across six benchmark datasets, RGA-Designer reported similar accuracy to ARG-Designer while using fewer tokens on average.