Machine Learning Faces Limits When Repeated Data Lose Identity
An arXiv preprint offers a formal framework for persistent units, but reports no tests on real data.
Search
2,451 results
961–980
Joint decoders show stronger agreement in synthetic tests
Preprint: A conditional generative method reported near-perfect observed-factor probes and stronger shared-output agreement in a controlled benchmark, while broader transfer remains untested.
Kestrel spots ML-KEM and ML-DSA in stripped binaries
This preprint reports that the detector matched ML-KEM and ML-DSA through tested build changes, but runtime-generated and XOR-rewritten tables escaped detection in the tested cases.
Preprint Tests a Multimodal Tennis Model With Nine Players
A study of nine collegiate players combined wearable data, questionnaires, fitness tests and match video to model wellness, injury risk, physical capability and playing style.
TiO2 simulations link laser intensity to film response
Preprint modeling links stronger fields to lower reflectance, higher absorption and fourth-power energy scaling in rutile.
Model suggests orbital angular momentum and helicity share a small-x power
An arXiv preprint reports a numerical calculation in a small-x QCD model, with a shared asymptotic power but ratios that remain sensitive to starting conditions.
New algebraic tests characterize hereditary QF-3+ rings
Preprint: The analysis connects stable and projective modules with injective envelopes and the maximal left ring of quotients.
Lattice study links small fermion modes with monopoles
Preprint: Numerical lattice results found that anomalously small Wilson-fermion modes generally matched monopole counts, with disagreements when a monopole and anti-monopole were one lattice spacing apart.
Simulations map valley polarization in monolayer MoS2
Preprint simulations suggest an attosecond probe could quantify valley polarization in monolayer MoS2, with a strong helicity-dependent signal in the model.
Agentic workflow passes 26 of 28 wafer-scale code tasks
An arXiv preprint reports higher executable pass rates and faster matched programs, but the test covered a specialized benchmark.
Fiber Alignment Alone Does Not Set Network Stiffness, Model Finds
Preprint: A model of two-dimensional fiber networks predicts that buckling can change densification and how far a cell-like contraction spreads.
Iron and Manganese Show Different Surface Responses in Lead-Tin Selenide
An arXiv preprint reports that submonolayer Fe and Mn deposits were associated with Rashba splitting on one crystal face and closer Dirac cones on another.
Text May Help Forecasts Only When Recent Values Are Not Enough
Preprint study finds that extra context matters most when recent history does not contain all the useful signal.
Flexible Smoothing Maps Score Higher in One Denoising Test
An arXiv preprint reports higher image-quality scores for a less regular map, while paired MRI tests showed little difference when maps were swapped.
Coupled Dirac valleys point to a new critical behavior
Preprint: The analysis reports a finite-coupling fixed point with broken Lorentz invariance, while the symmetry is asymptotically restored as the number of valleys grows.
Study finds three recurring shape patterns in normal corneas
Preprint: An analysis of 8,609 maps suggests three corneal shape categories that could guide future biosynthetic implant design.
Quantum-walk design reports lower model-based CNOT upper bounds
Preprint: An arXiv analysis compares decomposed and naive quantum-walk circuits, reporting lower model-based CNOT upper bounds in the stated examples.
SOPO-CD posts large speed gains in robotic placement tests
Preprint reports faster Tangram, Tetris and 3D bin-packing tests, plus an open-loop robot demonstration.
Preprint reports stronger attack resistance for a federated AI defense
A computational study links transferable attacks to differences between models and client data, then reports higher adversarial accuracy than federated adversarial training in tested settings.
Preprint Extends Path Abstraction to Expected Rewards in Finite Markov Models
The framework covers substochastic, discrete-time models and stores target-reaching probability with normalized expected reward for reachable absorbing targets.