Preprint reports 2-microsecond median in embedded scheduler test
An arXiv methods paper compares CIPS with a latency-aware controller in an embedded microbenchmark and a dynamic car-following benchmark.
Search
2,230 results
781–800
Linear model finds stable damping can still be slow
Preprint: A one-dimensional analysis finds state decay of 1/t and energy decay of 1/t^2, with both exponents sharp under localized damping.
Mathematicians link KdV wave variations through two spectral curves
Preprint: A formal study identifies branch points where several solution methods converge.
New method targets redshift with strongest dark-energy departure
Preprint: An illustrative analysis reports 2.78 standard deviations at the optimized redshift, versus 2.56 at the pivot.
Preprint finds limited gains beyond top-mass cache eviction
A computational analysis finds that physical memory, budget enforcement and question visibility can reshape comparisons between eviction methods.
Preprint reports near-zero bias for log-quantile fits in simulations
An arXiv v1 preprint reports exact one-step log-scale linearization, low variability in selected simulations and model-specific results in two observational examples.
Sparse Language Model Matches Dense Rival at Peak, but Is More Fragile
Preprint: A comparison found near-identical probing accuracy in a sparse and dense model, but much lower noise tolerance in the MoE system.
Preprint: Bidirectional Mamba model leads in simulated device trust
Across 500 simulated devices, the model posted the lowest reported error and highest modeled task value, but real-world networks were not tested.
Preprint reports sharp answers to two large-graph clique problems
The mathematical study gives exact equality cases for two conjectured bounds, but applies only to sufficiently large graphs and leaves its thresholds unspecified.
AI Models Put Justice First in Rare-Disease Ethics Test
An arXiv preprint found that equality-based choices made up the largest share of Justice selections, while decision-maker framing was strongly associated with the values chosen.
MVE Model Leads Rival Methods in Simulated Trust Tests
Preprint: An uncertainty-aware system for choosing collaborators led rival methods in a simulated distributed system, but the evidence remains limited to that test.
Preprint Links Spin Structures to Finite Geometric Measures
The framework adds Ramond-point cohomology and recursions, reports matching even and odd volumes in genus one and two, and marks broader equivalences as conjectural.
Preprint model posts 91.3% simulated collaborator-selection accuracy
TrustFormer reported lower trust-estimation error than LSTM and standard Transformer baselines, with the widest gap in complex simulated behavior.
Crowder size linked to shape and motion in polymer simulations
Preprint: Denser modeled crowding was associated with a more compact polymer, while the two solvent models showed contrasting diffusion and relaxation patterns.
Preprint reports matrix theory match for selected graviton amplitudes
The arXiv study links a supersymmetric state count to gravity results in one charge chamber and a special plane-wave configuration, with clear limits.
Preprint Finds Sharp Path-Deletion Threshold in Directed Graphs
A proof confirms the k = 1 case of Mader’s conjecture and derives new arc-deletion results under related degree conditions.
Most corporate code repositories lack committed AI practice
Preprint: A four-level model placed 66.7% of 441 repositories at Level 1, while none reached Level 4.
AI training method scores better on harder questions about new facts
Preprint: A three-stage reinforcement-learning recipe was reported to improve a language model's performance on multi-source and reasoning questions while keeping direct fact recall comparable.
Preprint Reports Error-Mitigation Results Vary by Noise
Numerical tests reported lower mean losses under uniform synthetic noise, while a heavy-hex case using hardware PTMs showed no reported improvement.
Badminton AI learns to adapt shots and recover in self-play
Preprint: In a physics-based badminton simulator, self-play agents changed shots and recovery with the opponent, but the human comparison was only a limited sanity check.