Model predicts a teleportation-like proton spin signal at 151 MeV
A theoretical preprint models entangled three-proton scattering and finds a narrow polarization window, but reports no experimental demonstration.
Search
2,451 results
621–640
Layered AI defenses stop tested jailbreaks but reject benign prompts
Preprint: A seven-layer test found no residual attacks in its run, but the stack refused 81% of benign prompts and offered no clear gain over its strongest layer.
Critical Gaussian Chaos on the Circle Has Vanishing Fourier Tails
An arXiv preprint reports an almost-sure theorem for the canonical critical chaos, while leaving the speed of decay unresolved.
Bangla medical system scores 0.95 on a crafted test set
Preprint reports strong test results, but contains no clinical validation.
A simple rule identifies which atomless distance laws can exist
Preprint: A theorem says an atomless distance distribution can be realized exactly when zero lies in its support.
Urban model suggests stronger links for elevated NTN terminals
Preprint: In synthetic urban scenes, the proposed analytical model produced higher simulated coupling gains than the current 3GPP model, while field validation remains to be done.
Proposed particle-physics distance bound passes one test, fails others
A 28 August 2026 arXiv preprint reports that the SharpDC rate bound holds at leading order in one supersymmetric large-N regime but fails in several other theoretical examples.
A new sequence encoding keeps time shifts exact in one test
Preprint: An arXiv study reports that its Sinusoid shift operation matched a 10-step shift exactly and ran about five times faster than standard FPE at one tested vector size.
New Theorem Extends a Classic Dynamical-Systems Result to Group Bundles
Preprint: The work extends the result under explicit topological assumptions and links group-bundle compactifications to spectra and dynamical systems.
Preprint sets conditions for smoother solutions to nonlocal equations
Preprint: A proof-based study says locally bounded solutions are locally Hölder continuous under structural assumptions.
Training Helps AI Match Medical Images Across Wavelengths
A preprint reports that cross-spectral training preserved comparable RGB performance while limiting failures on synthetic and modified benchmark images.
Preprint links denoising to lower 3D crop reconstruction error
An arXiv study reports lower real-data reconstruction error in denoised conditions, but its best model changed by dataset and by the score used.
Work function predicts hydrogen-evolution activity on Ag-Au-Pd-Pt alloys
Preprint: The coverage-corrected adsorption-only model scored 0.955, compared with 0.969 for the model combining adsorption energy and work function.
Mathematical proof links 2-sphere boundaries to Kleinian groups
Preprint: Under strong geometric assumptions, a new theoretical result extends a Bonk–Kleiner question to relatively hyperbolic groups.
New proof makes a key Coxeter group question decidable
This preprint reports a proof that links angle-compatible Coxeter generating sets through finitely many twists and a conjugation, and establishes an algorithmic decision procedure for Coxeter-group isomorphism.
Preprint sets out stability tests for Vlasov-Maxwell models
The paper presents mathematical conditions for stability, observer convergence and disturbance attenuation, while a reduced benchmark highlights important numerical caveats.
False Channel Reports Could Disrupt 6G NOMA Decisions
A preprint traces manipulated channel feedback through power allocation, scheduling, decoding, fairness and secrecy, but reports no measured attack effects.
GaussDB Posts Strong 30TB Score, But Results Are Unaudited
This preprint reports a composite TPC-H score 40% above Hologres at 30 TB, while the authors say the benchmark results had not been audited.
Math preprint finds a nerve that changes under thinification
Preprint: The Roberts–Street nerve is non-contractible in one formal setting but is contractible after every nonnegative thinification.
Study reports a more targeted way to select what language models forget
A preprint study reports higher forget-sample retrieval and model utility with GRACE, while quantitative forget quality changes little in the reported tests.