A theoretical preprint reports that fixed-order Khatri-Rao sketches can achieve a near-linear subspace-embedding dimension in the subspace size, improving the stated dependence on that size over earlier bounds. The result is a proof under sub-Gaussian assumptions, not an experimental performance claim.
A large-detector simulation predicts thousands of selected atmospheric-neutrino events over a long exposure, with the biggest tested spectral differences tied to the model of the initial neutrino-nucleus interaction.
A chirped femtosecond-laser system produced a measurable broadband heterodyne signal in a single-channel laboratory test, while projected multiplexed sensitivity remains untested.
A preprint reports a broad metal-organic-framework dataset and two released models that cut adsorption-enthalpy error while retaining comparable performance on selected tests.
A computational study reported faster convergence for a new graphene phonon-transport solver, with the biggest gains appearing as conventional iterations slowed at larger length scales.
A tiny two-recording test found that a parameter-free color extrapolator beat copying the last frame and every trained comparator on held-out copper in both directions. The air-to-chamber advantage was individually separated from zero, while the chamber-to-air interval included zero.
A critical review of lead-halide perovskites presents a testable framework linking their electronic structure, screening and carrier motion. It also warns that claims about universal defect tolerance, bulk ferroelectricity and dynamic Rashba effects remain unproven.
An evaluation in simulation and on a physical robot found that one shared interface could support three instruction modes, with gesture-based modes often performing better when wording, surfaces or objects changed.
A small observational study of blue field candidates confirmed nine ultra-diffuse galaxies and eight other low-surface-brightness galaxies. After an aperture correction, all targets were described as lying within the star-forming main sequence, although the corrected star-formation rates were treated as upper limits.
A laboratory preprint reports that photon emission from transmitted helium ions varied strongly with exit charge state, while differences among carbon, silicon and silicon carbide were small.
H-Scale uses calibration activations to choose hardware-valid NVFP4 scales, with higher average scores reported across Qwen3 and LLaMA tests but a narrow evaluation scope.
A laboratory study combines controlled microfluidic interface renewal with separate equilibrium and diffusion measurements to test whether dynamic tension can reveal adsorption rates. Its two surfactants required different model descriptions.
A version 1 arXiv preprint presents a conditional route for excluding non-trivial solutions of a generalized Fermat equation at sufficiently large prime exponents. Its imaginary quadratic example lists three relevant unit solutions and reports relative density 5/6 among squarefree positive integers.
Gibbs-family methods led the aggregate scores in Traffic Hourly, Electricity Hourly and Solar Weekly, but classical baselines remained on top in some M4 frequency and disagreement groups.
A mathematical preprint describes a new family of permutation group polynomials over finite fields of odd characteristic. It pairs the family with an explicit companion, distinguishes it from known families considered in the paper, and gives exact counts for the new and three specified earlier families.
A theorem-driven study gives coefficient, growth and function-space bounds for K-quasiconformal harmonic mappings, while leaving key sharpness questions open.
The preprint reports stable long-time propagation in most tested cases and favorable benchmark comparisons, alongside 908 runs that exploded in a 1,620-run stabilization sweep and a breakdown in the most nonlinear wall test.
A one-dimensional light-and-matter model shows a self-organized interface between a vacuum-like region and a chaotic bulk, with distinct scaling near critical points.
VICT uses a task's terminal verifier to trace credit through long action sequences instead of using it only as a final pass-or-fail signal. In ALFWorld and WebShop tests, it beat GRPO and achieved higher validation AUC over the same 300 updates, while the authors caution that verifier-defined links do not establish causal necessity.
The results point to a trade-off: the transformation embedding was more useful for distance-based comparisons, while the processed-audio embedding was more useful for recovering chain and parameter information.
An arXiv preprint proposes a 3D-grounded test for robot video models. Cosmos led composite scores, but detailed checks found weaker object localization and trajectory accuracy, underscoring the gap between plausible footage and executable behavior.
A laboratory study reports more than 40 dB of measured isolation from a switched-resonator CMOS circulator at a tuned operating point, along with circulation reversal in 0.6 nanoseconds.
A mathematical preprint derives a boundary observability estimate for critically singular wave equations on general domains. The theorem depends on convexity and a sufficiently long observation time, and the authors say it does not establish boundary controllability.
A theoretical result forces the number of cells in a bent partition to be a prime power and narrows its exponent using the dimension of the underlying finite space.
A mathematical preprint develops an unbiased idealized estimator and a two-stage randomized variant, then tests truncated versions on kernels and spectral filters.
A video AI method called Token-Budget Distillation retained strong benchmark scores after aggressive visual-token compression, although its training still depended on the uncompressed teacher model.
An anatomy-aware system called CheXtriev reported stronger case-retrieval scores than global and local comparison methods on selected chest radiographs. The gains were especially notable for several lower-prevalence findings.
A large Web of Science network analysis finds that China's high share of domestic citations is substantially explained by publication scale, while its relative home bias has declined and its research disruption has moved closer to Western levels.
An arXiv preprint tests a multi-critic training method for robots that push and transport objects and open a dishwasher through contact. It reports 94.1% simulation success and 69.0% success in 58 trials on four unseen objects, while the dishwasher test is qualitative.
A laboratory preprint reports that light may leave optically addressable traps in BaF2, with changes in luminescence yield and a spatial readout of excitation sites.
Scientific publication4 min read
Showing 30 of 2044 articles
Research fields
Explore science by topic
Start broad, then move into one of thirty focused subject desks.