Preprint says round disk is hardest convex shape to map onto a plane
A geometric proof says equal-area convex regions on a sphere can be mapped to the plane within the round disk’s distortion benchmark.
Search
418 results
261–280
Preprint argues Spatial AI should work with spatial statistics
The conceptual review says spatial scale, dependence, sampling design and uncertainty must remain central as models become more flexible.
Preprint reports a universal bound for Riesz transforms on stratified groups
A proof-based analysis puts the real-valued weak-type constant at 2 across the full class, while leaving several questions open.
Preprint finds finite high-order Drude weights where theory predicted divergence
Exact Bethe-ansatz calculations in a one-dimensional Hubbard model challenge a low-energy prediction at quarter filling, while half-filled odd-order weights show exponential finite-size suppression.
Preprint builds a chain in which both π2 maps are zero
The formal construction links three two-complexes, while a separate Lie-ring example shows a nonzero identity among relations in the smaller presentation.
Preprint: Targeted Vibrations Linked to Different Proton-Transfer Outcomes
Computer simulations of two fluorophore models found ultrafast single- and double-proton transfer under selected vibrational conditions, while room-temperature baseline runs showed little transfer.
Preprint finds a constructed operator can pass a calculus test yet fail stochastic regularity
The theoretical analysis maps exact deterministic and stochastic conditions for diagonal operators and ties them to the geometry of Banach spaces.
Preprint finds a second critical boundary in a model of learning from measurements
Calculations and finite-size simulations point to a three-phase meeting point, with a possible connection to toric-code error correction.
Preprint reports stronger jump-control results with active suspension
In selected simulations, a combined controller passed a post-landing attitude test in all 24 boundary runs, but the findings remain tied to specific ramps, speeds and a modeled vehicle.
Preprint tests a neural-network method for shaping 3D light patterns
In simulations and fluorescent-bead experiments, the approach reproduced several designed point-spread functions and used fewer reported iterations than pixel-based methods in one benchmark.
Preprint Reports Higher Simulated Secrecy Rates With a Movable Drone-Mounted Reflector
A mathematical model reported higher rates when the reflector’s position, orientation and reflection settings were optimized together, but hardware and field performance remain untested.
Preprint finds conditions for random-walk samplers to handle steep targets
A mathematical analysis derives a positive acceptance bound across states and dimensions, but does not test the method on real data.
Preprint reports collapse suppression in a singular quantum model
The theoretical study reports different outcomes for cubic and quintic self-repulsion, along with localized states in steep outward-pushing potentials. No experiment is included.
Preprint Finds LFU Hard to Beat in Large-Language-Model Caches
Across three corpora, three cache capacities and two embedding encoders, no tested policy materially outperformed least-frequently-used replacement; a quality audit sharply reduced apparent hit rates on two workloads.
Preprint: Staged AI training scores higher than direct fine-tuning in most tests
The IAR method showed stronger document-question and general-benchmark scores in seven of eight main settings, but every checkpoint came from a single training run.
Preprint: Fast networks were only part of the picture in Earth-observation tests
An on-premises system built around a 100 GbE fabric moved data at high aggregate speeds, while measured throughput varied with endpoint hardware and remote network paths.
Preprint links Floer theory with free-loop spaces
A mathematical paper presents MU-module and framed sphere-spectrum equivalences for cotangent bundles, while finding that sphere-level base-change does not generally recover the MU result.
Preprint strengthens a rigorous bound on 1324-avoiding permutations
A new construction puts the certified growth-rate lower bound at 10.617, while leaving the exact rate unresolved.
Preprint: Surgical AI model performs better when it forecasts future scenes in chunks
In robot-assisted prostatectomy videos, chunked forecasts also kept predicted instrument paths closer to the benchmark, although long-range errors remained.
Preprint finds sleep-signal maps differ by sex and age
A model of 105 clinical HSAT recordings found recurring links alongside subgroup differences, but the observational results are not a clinical rule.