Reinforcement learning shows lower errors in quantum simulations
Preprint: A simulated protocol used energy signals to choose each step and reported lower errors in larger tests, but no quantum processor was involved.
Search
2,451 results
401–420
AI models explain theoretical computer science better than they can prove it
Preprint: A benchmark across 143 research papers found formalization was the weakest stage, and only six of 64 generated claims survived expert review and proof checking.
Expanding Regular Graphs Can Still Miss Most Vertices in a Cycle
Preprint: Two explicit graph families have small longest cycles despite their expansion properties, while the robust-expander case remains open.
Every repetition level appears in infinite words using three symbols
Preprint: The refined exponent spans every value from 1 to infinity over three symbols, while it equals 1 almost surely under uniform Bernoulli measure.
Stack Framework Links Rigid Cohomology and Arithmetic D-Modules
Preprint: A mathematical construction links arithmetic de Rham stacks with rigid cohomology, isocrystals and arithmetic D-modules.
Preprint Maps Which Causal Formulas Keep Denominators Stable
Analysis finds exact Gaussian calibration for some covariance formulas, while others become unstable near weak identification.
A six-valued logic gets a probability framework for gaps and gluts
Preprint: A formal construction tracks belief, conflict, uncertainty and reliability, and shows when two styles of updating agree.
Math Study Maps the Geometry Behind Stochastic Regularity
Preprint: A theorem-based analysis gives exact tests and shows that a strong operator calculus can coexist with failure of stochastic maximal regularity.
New analysis sharpens the cost of fair sampling from biased sources
Preprint: The Bernard-Letac construction is reported to be exactly fair, with a prime-case cost formula and lower mean draws in tested composite cases.
Banda Aceh study finds no clear inflation or rate-spread link
A 15-year time-series analysis found a long-term level relationship, but neither inflation nor interest-rate spread estimates were statistically significant.
Mathematical study finds uniform Riesz-transform bounds stop at p = 2
Preprint: A proof-based analysis finds group-independent bounds through p = 2, but no single finite bound works across all groups for p > 2.
Preprint raises lower bound for 1324-avoiding permutations
An arXiv version 3 paper reports 10.412263, up from the previous rigorous lower bound of 10.271012, after counting a two-direction staircase construction.
Preprint reports lower 3D hand-tracking errors during occlusion
A diffusion-based method had lower errors on most reported benchmark measures, but its out-of-sight results remain wrist-aligned and the system works offline.
Preprint maps the rules for adding mathematical orders
This arXiv preprint reports an affirmative answer to Tarski's finite-coefficient problem and a classification of commutative semigroups represented through ordered sums.
Ion-channel study finds response memory beyond dwell correlations
Preprint: A mathematical framework and BK-channel analysis suggest that response-relevant memory can differ from ordinary measures of dwell correlation.
Simulations suggest a route to loading one molecule per trap
Preprint: A modeled NaCs protocol predicts more than 99% single-site isolation and 95% loading per tweezer across an array.
WZW analysis finds a sharp ceiling for torus topological entanglement
This preprint finds that finite stacking preserves the leading coefficient, but it covers only a defined class of mathematical theories.
Brain cells handle cortical and thalamic signals differently
Preprint: Two long-range inputs engaged cholinergic interneurons in the nucleus accumbens core in sharply contrasting ways.
Review warns against overreading Mendelian randomisation in rheumatology
A clinically focused review says the method can sharpen causal questions, but genetic signals are not automatic substitutes for treatment evidence.
Many Axial Spondyloarthritis Patients Reached Remission, but Often Lost It
A peer-reviewed German cohort found that 58.5% reached the study's remission threshold at least once, but remission was often not maintained.