New dataset tests whether AI can diagnose through conversation
Preprint: MTDiag combines synthetic cases, hospital records and case reports, but does not yet compare model performance.
Search
2,230 results
661–680
Memory-augmented decoder leads when neural data are scarce
Preprint tests found VN-SST ahead of a Transformer on three public neural recordings, with the widest gap under limited training data.
Preprint finds matrix factorization counts grow much faster for zero targets
This arXiv preprint gives a full asymptotic for fixed nonzero-determinant targets in dimensions 2 and above, a uniform upper bound, and a separate formula for the zero matrix.
Model spots car groups and unusual cases despite missing data
Preprint: A statistical framework combines clustering, regression and missing-data treatment in an Automobile dataset analysis.
Mathematical test links positivity to oscillator-semigroup operators
Preprint: A proof in the complexified Schrödinger representation finds an exact link between a matrix property and bounded intertwiners.
Preprint study reports charm ratios vary with jets and event activity
Using Pythia simulations and selected ALICE measurements at 13 TeV, the authors report different baryon and meson patterns under jet and event-activity selections.
Sentence-level scores may make language-model comparisons fairer
A preprint analysis finds Sentence NLL is less tied to surface text counts, but the metric is not a perfect measure of model quality.
Alkali-intercalated WTe2 shows a new structure and narrow gaps
Preprint: Potassium, rubidium and cesium versions showed wider spaces between WTe2 layers, while transport testing focused on the rubidium phase.
Te-Doped PtBi2 Shows Bulk Superconductivity Near 2.4 K
Preprint: A laboratory study reports a centrosymmetric structure and near-complete superconducting volume in selected crystals, while point contacts show higher, variable transition temperatures.
Solver reaches near-complete MaxCut coverage in 0.9 seconds
Preprint: A GPU version nearly matched the reference hypervolume on two tested instances, while refined results reached 100% on both.
Mathematical study sets error bounds for noising and denoising
An arXiv preprint gives asymptotic bounds for a path construction that mimics forward diffusion and approximates time reversal, with results limited to specified mathematical settings.
Power-grid neural network reports low violations across five feeders
This preprint tests a self-supervised method on five simulated IEEE feeders, from a 13-bus model to an 8500-node system, as loads, distributed energy resources and switch settings changed.
MO-IKE Posts Higher Frozen-Model Scores, but Retention Lags
Preprint: MO-IKE records higher scores than comparison methods in several frozen-model tests, while its UniEdit row is much lower.
ROS2 Connect Records Lowest Latency in a Single-WAN Test
Preprint: In one real-WAN benchmark, ROS2 Connect reported lower latency and steadier timing than three alternatives.
One 3-D Ising conformal mismatch was lower at higher derivative order
Preprint: A 3-D Ising calculation finds a smaller tracked violation near a whole-field regulator choice, while offering no formal convergence proof.
Best AI Model Gets 33.7% of Olympiad Physics Problems Right
Preprint: An arXiv study tested 18 models on 11,586 bilingual problems and found a wide gap between machine and student performance.
Energy Per AI Token Rose More With Long Contexts in MHA Models
Preprint: In four open-source models, energy growth was much larger for MHA than for GQA-based configurations, while batching was associated with more than 80% lower energy per token in nearly all tests.
Adaptive AI Teams Find a Middle Ground on Speed and Token Cost
Preprint: A model and live deployments suggest adaptive communication can sit between ring-only and complete-only strategies, while its guarantees depend on strict assumptions.
Neutrino software reproduces ACHILLES prediction shapes in tests
Preprint: ADoNIS reports agreement with ACHILLES and tests gradients in fitting, sampling and detector-response unfolding.
Math preprint maps when a fractional equation exists or blows up
A theorem-based analysis links solution behavior to initial energy, sign conditions and the dimension of the model.