Study predicts tiny rates for rare top-quark meson decays
Preprint calculates tiny Standard-Model rates and an almost absent right-handed vector signal.
Search
2,451 results
1261–1280
Welding cable model changes how robot joint loads are calculated
Preprint: A planar simulation found that a modeled welding umbilical can noticeably alter a robot’s dynamic response despite its much lower mass.
ClueWeaver Leads Local Methods on Four Evaluated Narrative Benchmarks
A preprint reports that ClueWeaver recorded 59.0% overall accuracy across four long-context narrative benchmarks, with paragraph-level evidence traces.
Near-Field Beamforming Nearly Matches Full SVD at Lower Cost
This arXiv preprint reports near-full-SVD rates and much lower CPU time for a geometry-based dual-UPA design.
FedRep records 1.34 bps/Hz in a simulated three-cell comparison
This preprint reports simulated results for a system that jointly optimizes beamforming and fluid-antenna port selection across multiple cells.
New Software Model Targets Reversible Hot-Swapping of Components
Preprint sets out formal rules for reversing component effects and managing changing dependencies, with a qualitative Koishi case study.
LY UMa Appears Bluer When It Is Brighter, Preprint Says
Preprint: Optical measurements estimate a K7-K8V donor, but leave LY UMa's classification unresolved.
Simulated quantum reservoirs show a scaling edge as information scrambles
Preprint simulations found the strongest short-term forecasting and superextensive memory scaling near the onset of information scrambling.
OpenVeinNet reports lower error in held-out finger-vein tests
Preprint: An arXiv version 1 study reports the lowest error in four of five full-subject splits, while few genuine comparisons limited resolution at the strictest VERA threshold.
New framework gives sparse bounds in abstract Orlicz-Morrey spaces
Preprint: Theoretical results cover weighted bounds, commutators and duality, but depend on structural assumptions.
Game-building traces linked to stronger world-model test results
Preprint: A proposed system blends human acceptance feedback with game-engine checks and reports leading Unity scores plus positive transfer and embodied diagnostic results.
Preprint reports lower per-step time in neural-network tests
An arXiv preprint reports less time per iteration than classical and Krylov-subspace Levenberg-Marquardt methods in three synthetic network comparisons, while final training errors were similar.
Bird-recognition model reports 87.23% accuracy with audio and video
Preprint: ACF-Net was tested on BirdPro, a 194-class benchmark, and reported the strongest score in a test with deliberately mismatched training audio.
Study finds search scores hinge on how AI expansions are blended
Preprint: A controlled benchmark found that changing how the original query and generated text were combined improved retrieval scores and reduced index-side request costs.
Vesicle simulations find membrane models predict different shapes and motion
Preprint: A coupled fluid and membrane model matches reference shapes, but its predictions for rotation and deformation diverge at lower reduced volumes.
Plasma model predicts discrete waves in a helical magnetic mirror
Preprint: A numerical model predicts coupled ion-acoustic modes and fast-flow instability, while leaving the proposed ion-scattering link untested.
Preprint reports shorter delays in multi-agent AI workflows with TOPAS
Tests on one A100 GPU found lower task delays than selected scheduling baselines, but the evidence is limited to the configurations studied.
Temperature gaps may shape results from electrically heated tests
Preprint review says direct-current testing can reveal material behaviour, but only when temperature fields and other uncertainties are controlled.
Preprint compiler reports close matches and large FPGA speedups
Syn2Logic matched selected neural models and reported major FPGA gains in a 302-neuron C. elegans test, alongside separate Sudoku and MNIST results.
Video benchmark finds AI models miss complex instructions
An arXiv preprint finds that multimodal models often meet part of a video prompt while failing to satisfy all of its requirements.