761–780
A peer-reviewed laboratory study matched programmed dome and saddle curvatures and reported high fibroblast viability while transformed structures retained integrity for up to 15 days in culture.
A peer-reviewed study reports an association between the oxide film formed before sulfurization and the coverage, defects and electrical output of flexible MoS2 devices.
Preprint: A method that checks behavior-relevant tasks before accepting changes reported higher held-out scores, but its budget comparison was not compute-matched.
Preprint reports faster robot action decoding with similar or higher task success in simulations and physical tests, but notes a cold-start cost and successful-trial-only timing.
Preprint: A prompting method transfers failure patterns from weaker models, reporting modest gains with fewer target-model tokens.
Preprint: Simulations outline possible kaon yields and physics reach for a proposed Fermilab rare-kaon program.
Preprint study finds optimization-based methods were strongest overall, but case-to-case performance remained too variable for unmonitored dose accumulation.
Preprint: A year-long subarctic field campaign found simple proprioceptive odometry stayed comparatively robust as sensor methods degraded.
The two-dimensional classification separates real and complex matrix cases, while a Jordan-block result needs checking.
Preprint reports complexity guarantees and stronger numerical results for a method that changes penalty growth as constraints improve.
Preprint analysis of RHIC collisions finds stronger initial suppression of strange than light quarks, but only weak evidence for slower strange equilibration.
Preprint: A review found none of 2,811 selected health-themed descriptions met all eight mandatory HealthDCAT-AP Release 7 properties.
Preprint: A review says some protein positions face functional constraints beyond the need to stay folded.
Preprint: A five-dataset simulation found that the strongest raw scores often came with higher circuit costs or weaker generalization.
An arXiv preprint reports that PMFRec led nearly all cold-item comparisons on CiteULike and XING, with gains of up to 7.56% over the strongest baseline.
Preprint: KLOD reported highly reliable edits and strong preservation of unrelated predictions, while its threshold exposed a tradeoff with rephrased prompts.
Preprint: A large English-language test found that AI-edited passages were more often mistaken for human than for fully AI-generated text.
An arXiv preprint reports stronger benchmark averages and steadier rollouts for SpikeOPD, but the evidence comes from computational tests, not hardware measurements or users.
An arXiv preprint reports better depth and segmentation scores after attaching Distortion Extenders to frozen perspective-trained models.
Preprint: A Qwen3.5-9B model that generates complete HTML/CSS documents for e-commerce creatives scored 94.0 automatically and 90.0 in blind expert review.