Robot-learning method reports higher success on held-out tasks
Preprint: RA-VLA retrieves action-aware robot demonstrations during evaluation and reports higher average success on simulated LIBERO and real-world UR5e tests.
Search
2,111 results
921–940
Models suggest dark-matter halos could weaken neutron-star lensing
A preprint reports lower modeled deflection inside extended halos, with the predicted effect far below current direct-imaging resolution.
Model finds directional volatility links among UK firms
This preprint reports a model that learned a sparse, directional volatility network for 20 UK-listed firms and had lower errors than preset networks.
Satellite Model Finds Coherent Links Could Reach More Power
Preprint modeling compares coordinated satellite links and finds quadratic received-power scaling under coherent coordination, but the result remains theoretical.
Local Fairness Can Demand Much Larger Clustering Slack
Preprint: A study of fair hierarchical clustering finds that local checks can require much larger slack across a full hierarchy.
NUCLEUS detector records 2.16 eV resolution in lab
This preprint reports electron-volt-scale resolution and stable operation in six cryogenic detectors, while calibration and reactor-site backgrounds remain open questions.
Smaller echo-control model reports similar results with 70% fewer parameters
Preprint analysis reports about the same benchmark performance from a smaller model at 14% of a medium model's computational complexity.
Task-shaped AI harnesses linked to higher scores and lower costs
Preprint: The study reports higher scores and lower costs for task-specific harnesses generated by a learned meta-agent across nine benchmarks and several model families.
Model links quantum neutrino dynamics to possible gravitational waves
This preprint reports a theoretical scan that finds a model-dependent correlation, not a one-to-one way to infer gravitational-wave observables from neutrino parameters.
AI screen flags 97 possible room-temperature electrolytes
Preprint: A staged computational screen cut 30,364,908 generated structures to 97 predicted candidates, but laboratory testing is still needed.
Painlevé symmetries leave a finite trace on monodromy surfaces
An arXiv preprint examines four Painlevé equations and finds their surface monodromy matches finite Weyl groups.
Preprint proposes building a sensing network from many devices
A conceptual paper says correlated observations from distributed UEs could support target detection, tracking and environmental perception.
Graph Detection Scores Higher Than Sequence Decoding on CubiCasa5K Test Scans
Preprint reports graph detection ahead on 296 corrected CubiCasa5K test scans, while sequence decoding leads on small plans, an open-plan tier and finetuned ResPlan data.
Cobalt-doped films show stronger magneto-optical response with little added loss
Preprint: Co-doped cerium oxide films showed a much larger Faraday signal in one comparison, while the proposed microscopic explanation remains unconfirmed.
Stablecoin Defense Test Catches Most Simulated Attacks, With False Alarms
An arXiv preprint reports high recall and millisecond screening in simulated tests, but lower precision leaves live-chain performance untested.
New model scores higher at choosing camera views
An arXiv preprint reports stronger benchmark results and tests how target-video data affects camera-view recommendations.
Model Finds Local Demand Can Reinforce Generative-AI Automation
Preprint: A theoretical local-service economy can tip toward automation or human augmentation as firms revise their expectations.
New framework treats Banach-space integration as a functor
Preprint proposes a categorical approach with strong exactness results, but finds that infinite sheaf gluing can still fail.
Preprint narrows contravariantly finite resolving subcategories to three
A mathematical preprint reports that, over a henselian local ring, only three forms can occur under the stated approximation condition.
Preprint: Model Posts Lower Errors in Typhoon Gust Forecasts
WDANet outperformed the tested machine-learning models and ECMWF-HRES at short lead times, while ECMWF-HRES was stronger later.