Preprint reports AI model leading listed methods on an image-and-text sarcasm benchmark
The system reported 92.62% accuracy on MMSD and 89.66% on MMSD2.0, but the study offers no formal uncertainty estimates.
Search
2,451 results
2081–2100
Preprint reports lower MILP gaps using early-solution predictions
EnCore had lower final gaps in 11 of 12 Gurobi settings and transferred to SCIP without retraining, but the evidence comes from computational benchmarks.
Preprint: Bipedal robot changes its stride as ground softens
MILD combines terrain modeling and reinforcement learning; tests on an EngineAI SA01 reported lower transport costs and maximum torque than two comparison controllers.
Preprint: Cypress testing was faster, but not flawless, in a single web-app study
A 27-case evaluation found shorter automated runs and zero recorded repair time for a data-cy locator suite, while two interactive tests fluctuated between passing and failing.
Preprint finds strict limits for fractional revival in directed graphs
A theoretical analysis rules out the effect in a class of abelian Cayley graphs while identifying necessary and sufficient conditions for it between opposite sides of semi-Cayley graphs.
Preprint reports low-background setup for cryogenic magnetoelectric measurements
The closed-cycle system tracked magnetic-field-related voltage and dielectric behavior from 20 to 300 K, while reference tests helped separate sample response from instrument pickup.
Preprint reports mixed results from a model of a charged collapsing star
The selected solution combines positive density and radial pressure with negative tangential pressure, a central singularity and a sign-changing energy diagnostic.
Preprint reports stronger cooperative-driving results with lower communication cost
G-MARK is reported to outperform V2V-GoT on eight of nine benchmark entries, with the largest gains on tasks requiring explicit cooperative evidence.
Preprint: Domain-Tuned Model Leads Russian 1C Code Search Test
A fine-tuned bi-encoder ranked first on an open 3,413-query benchmark, while an overlap audit produced similar clean-subset scores.
Preprint model tops reported 3D vessel-mapping benchmarks
A single flow-matching system led key scores across three public vascular-imaging datasets, but the evidence remains a computational comparison rather than a test of patient benefit.
Preprint Finds Single-Sentence Voice Fakes Hard to Spot
In a test of 82 IT professionals, selected ElevenLabs recordings were much harder to identify than recordings from two older speech-synthesis tools.
Preprint sets conditions for minimum-energy partitions of graphs
A mathematical analysis covers weighted finite and countably infinite graphs, while ladder examples show how boundary rules and weights affect whether minima exist and which shapes attain them.
Preprint sets out two explicit models for complex elliptic surfaces
The theoretical study develops double-cover and determinantal descriptions, then applies them to Halphen and Enriques surfaces.
AI compliance agents made rejected and rule-breach attempts in simulations, preprint finds
Tests found that plausible compliance language did not guarantee rule-grounded actions, while monitor scores depended on the evidence supplied.
Preprint reports microscope that tracks ultrafast changes in molybdenum disulfide
The 80-MHz system used photonic-crystal-fibre supercontinuum probes and reached a pump-defined spatial resolution of 0.51 micrometres.
Preprint: Robot-Vision Framework Scores 89.29% in Simulated Assembly Test
PVRA scored above the study’s ICP comparisons and FoundationPose with PVRA-predicted masks, but below FoundationPose with ground-truth masks.
Preprint reports quadruped robot jumping through narrow gates
A two-level learning system combined continuous movement skills with vision-informed velocity commands in simulation and on a Unitree Aliengo robot.
Preprint: X-ray snapshots track hidden casein changes as milk gels
A time-resolved lab study found hydrolysis preceded aggregation, while the changing scattering pattern fit a porous “snowball” model better than a diffuse core-shell model.
Preprint proposes a local test for harmonic stability in power systems
In one averaged converter model, HSS eigenvalue indicators stayed negative and a matrix-based stability condition was feasible across the tested truncation orders.
Preprint reports proof of a coding-theory distance conjecture
A refined Bézout bound also yields lower bounds for generalized Hamming weights, while a broader Cartesian-code conjecture remains open.