Model suggests a new route to tunable light-hole qubits
A preprint reports modeled light-hole qubits with stronger spin response and a bilayer that switches between heavy-hole and light-hole character.
Search
2,230 results
1301–1320
OFDM preprint finds compact model handles amplifier distortion
Preprint: Simulations recovered nonlinear channel behavior at low signal-to-noise levels and supported compensation close to an ideal linear system.
Human reviewers favor a rebuilt benchmark for key-point analysis
An arXiv preprint reports that human reviewers preferred re-annotated structures for grouping, key-point quality and coverage over structures derived from ArgKP21.
Neural model speeds porous-media tests, but high-contrast accuracy lags
This arXiv preprint reports up to 16.3-fold faster online calculations in lower-contrast two-dimensional tests, while high-contrast results did not reliably match the standard method.
Diffusion-based detector flags backdoored vision models in tests
Preprint: DEFUSE reported AUPRC scores from 0.81 to 0.96 across data-poisoning attacks and also tested other image domains.
A Robot Grasped Objects With Partial 3D Reconstructions in Lab Tests
Preprint reports 65.6% success across 192 grasp attempts, while a low-resolution method used fewer views than a standard approach.
Protein model benchmark scores differ across sampling setups
A preprint compares three inference protocols for protein models and reports different benchmark figures alongside the runtime cost of guidance and search.
Finite-Ring Family Breaks a Conjecture on Polynomial Null Ideals
Preprint: The paper proves two-sidedness when the Jacobson radical has nilpotency at most 3, but leaves the smallest counterexample open.
Simulations suggest gas disks may form around black holes in UCDs
Preprint simulations examine whether AGB gas can gather into disks around black holes in ultra-compact dwarf galaxies, and how modeled conditions change the outcome.
Dense-matter instability maps differ under TRS, nearly overlap under MSS
An arXiv preprint compares three ultraviolet prescriptions in a two-flavor Nambu-Jona-Lasinio model and reports near-overlap under medium separation.
FedAvg and SCAFFOLD Show Different Results in Sharpness Tests
This preprint reports a slight FedAvg loss edge on CIFAR-10 when Edge-of-Stability dynamics appeared, while SCAFFOLD led on MNIST.
MA-VLA reports higher scores on specified multi-arm collaboration patterns
Preprint: MA-VLA reported higher benchmark scores than Pi0 and nonzero results on specified collaboration patterns, with tests limited to selected simulations and one dual-arm platform.
Euler analysis finds smooth flexibility, analytic rigidity
Preprint: The analysis finds non-shear smooth steady states under a stated profile condition, while a dense family of analytic profiles permits only shear states.
Action-Aware Model Leads E-Commerce Forecasting Tests
Preprint reports lower offline errors and higher reported merchant LTV and ROI, but the online comparison lacks key design details.
Mathematical study finds balls minimize a key PDE constant
An arXiv version-1 preprint reports a universal lower bound and sharper results for smooth domains, ellipsoids and annuli.
Model predicts fluffy PET clusters and many smaller particles
A preprint models what may remain after enzymatic plastic degradation, including release times for bottle and textile PET.
Exoplanet Spectrograph Meets Key Stability Test, Early Results Show
Preprint: RISTRETTO’s seven-spaxel instrument covered 617 to 860 nanometres and produced first spectra on sky, but calibration and alignment work remains.
Preprint rules out nontrivial outgoing scalar-wave modes in Kerr-de Sitter
A mathematical proof covers the full subextremal range, while extremal cases and negative-imaginary-frequency modes remain outside its scope.
New four-manifold construction sharpens a key mathematical bound
Preprint presents an infinite family of spin symplectic spaces and a conditional route to the value 8.
Prior scores tied to shifts in AI judges' ratings, tests find
A preprint reports lower numerical scores and altered classifications under anchored metadata, while two tested prompts showed mixed or no clear protection.