Adaptive odometry reports low drift in degraded sensor tests
Preprint: A multi-platform odometry and sensor-fusion evaluation reports lower errors and online adaptation, while calibration and generalization gaps remain.
Search
2,451 results
1881–1900
Preprint Reports Stronger Robot Results With Visual Contact Cues
An arXiv version-1 preprint reports high success in grasping, cap unscrewing and calligraphy, but the tests covered one robot setup.
One-shot AI guidance outperforms static help in program searches
A preprint reports higher solve counts across several synthesis benchmarks, while the method remains untested on human or production coding.
A simple AI model segmented stroke lesions on MRI in seconds
Preprint: A DWI-only nnU-Net model was tested on three retrospective MRI datasets, but not on a fully unseen external cohort.
AI training system reports faster updates on agent workloads
Preprint: A 2026 arXiv paper reports higher throughput, lower memory use and stronger scaling in four benchmark workloads; its source code is forthcoming.
New Logic Links Reachability, Shape and Reversible Motion
Preprint: A formal study reports sound and complete systems for topological, finite, Alexandroff and polyhedral models with invertible dynamics.
Study finds most AI creativity metrics diverge from human ratings
This preprint reports weak links between seven automated metrics and human ratings in 200 stories, while an LLM judge scored AI-generated and human-generated stories differently across all 11 dimensions.
MIMONet reports better scores in most salient-object comparisons
Preprint: A three-resolution model was compared with 21 recent methods across six named image datasets, using only original-size saliency maps for final evaluation.
Preprint Method Keeps Stochastic Simulations Exact Forward
Tests reported similar accuracy to GS-ST across several benchmarks and a 3.0-fold reduction in the epochs needed to cross an oscillator error threshold.
X-ray study finds opposite strain in a switching GaN transistor
Preprint: High-speed imaging found compression and stretching in different regions of one industrial device, while simulations broadly tracked the timing.
Sterile Neutrinos Lead a Principle-Based Dark Matter Ranking
Preprint: A conceptual comparison puts sterile neutrinos first among four mainstream candidates by the paper’s standards.
AI method reports benchmark gains over GRPO with far less compute
Preprint: SRPO uses short reflections to create token-level training signals for language models tackling long-horizon reasoning and agent tasks.
Mathematicians extend wall-crossing formula to Calabi–Yau four theories
This arXiv preprint builds framing-independent semistable invariants and extends the formula to fixed-determinant stable pairs, subject to explicit assumptions.
Study derives exact formulas for how massless fermions share information
A preprint presents exact mode formulas for spherical regions, checks them on an 8,000-site lattice and treats a four-dimensional Rarita–Schwinger field.
Simulated Beam Shaper Shows Limits at Higher Super-Gaussian Orders
Preprint: A modeled diffractive optical element reproduced a lower-order profile more closely, while manufacturing and wavefront errors degraded the result.
BlueBird 6 Satellite Fainter Than Earlier Spacecraft
Preprint analysis finds the first Block 2 satellite was dimmer than Block 1 craft, even after correcting for distance.
AI agents report five math results judged novel in test
Preprint says an open-world system beat AlphaEvolve on three other problems in a 14-problem evaluation.
Few-shot AI method reports strong scores on industrial defect tests
An arXiv preprint describes DriftAD, which adapts text descriptions to local visual features and reports results on MVTec-AD and VisA.
GAP-Prompt reports higher accuracy on three image benchmarks
Preprint: GAP-Prompt reports results across three image benchmarks, while task-querying accuracy ranges from 44.30% to 55.75%.
OrbitNet leads Starlink tests with 3.34-metre average error
Preprint: OrbitNet posted the lowest mean absolute error on all six unseen constellations, although its other error scores varied.