Data-only AI clustering beats mean prediction in synthetic tests
Preprint: In synthetic tests, the method came closer to one of several valid outputs than random labels, but it remained behind a benchmark using true modality identities.
Search
2,111 results
821–840
GPU Kernel Study Reports Up to Twofold Gains in Matrix Solves
Preprint testing on three accelerators reports peak speedups of 2.05× over cuBLAS and 2.06× over rocBLAS.
Preprint model tracks tape temperature during dynamic process changes
The simulation reported a 4.02°C RMSE against measurements, but its parameters were tuned on the same trajectory used for testing.
Preprint method shows sharper phantom images while learning sound-speed corrections
Testing found sharper phantom images and visible vessel structure at 12.5% detector retention, while validation remained limited to simulated and physical phantoms.
AI Agent Design Separates Audit Trails From Raw Data Sharing
Preprint: An arXiv paper outlines field-level controls for limiting identifying details, while tests and a reported failure expose important trust gaps.
FedQoS stays close to centralized prediction in simulation
Preprint: In a physics-based campus simulation, FedQoS remained close to centralized prediction in two non-IID conditions, while access-selection results differed between mild and severe tests.
Offline AI shows gains in simulated multi-drone fleet tests
Preprint: An arXiv version-1 study dated 26 Aug 2026 reports higher return, communication and sensing measures, plus lower collision-risk counts in modeled multi-UAV tests.
Neural Network Builds Compact Formulas for Quantum Energy Levels
Preprint: In spin-chain tests, the model made compact formulas but did not beat a known quadratic basis on accuracy.
AI dataset method tops three fine-grained image benchmarks
Preprint: DeCO had the highest reported accuracy at low image budgets, but the results lack formal uncertainty tests and no public code was reported at submission time.
Underwater image scores can rise as 3D surfaces deteriorate
An arXiv preprint comparing five systems found image scores diverged from geometry measures in murky water and on an operational survey.
Preprint: Qubit readout reaches 97 ns in an operational test
A single-transmon test paired a 58-nanosecond measurement with 0.17(1)% assignment error and mean leakage of 2.7(2) × 10^-5 per measurement.
ATLAS finds tZq top-quark production in line with the Standard Model
Preprint: ATLAS reports its most precise tZq study yet, with cross-section measurements, spin information and limits on nine dimension-six coefficients.
Open-source tool tests Vietnamese text for signs of AI writing
A preprint describes a Vietnamese detector that accepts scanned and long documents, while its authors say human review remains essential.
Video model brings 3D motion and camera control to live generation
A preprint reports more than 20 frames per second at 480p, while also documenting blur and unexplained object changes in generated scenes.
Model Finds Exciton-Phonon Coupling Depends on State Parity
A preprint reports that same-parity elastic transitions cancel the long-range divergence, while opposite-parity inelastic transitions retain finite coupling.
Battery leaching model overpredicts recovery with peroxide at high acid
Preprint: A computational model of lithium cobalt oxide leaching matched lower-acid comparisons reasonably well, but overpredicted recovery when peroxide was present at 2.5 M hydrochloric acid.
Cross-layer AI retrieval system leads long-story tests
An arXiv preprint reports the highest displayed scores for PonsRAG, but the evaluation covered only long narrative reasoning benchmarks.
A CLIP adaptation reaches 99% on a sparse-label remote-sensing test
This preprint reports a class-balanced method that adds pseudo-labels to sparse remote-sensing training and outperforms listed baselines on two benchmarks, but its performance beyond them is untested.
Preprint Reports Graph-Coloring Girth Requirement Reduced From 11 to 5
A mathematical preprint reports conditional spectral-gap lower bounds and mixing-time upper bounds for proper colorings on finite graphs with girth at least five.
Wealthier groups tend to have richer support networks, study finds
Preprint: A 46-community analysis links wealth to support ties and wealth-based network separation, but cannot establish cause and effect.