AI system HypoForge tops reported generation and test-pass scores
An arXiv preprint reports results from a 13-task computational benchmark; POPPER recorded higher execution success.
1581–1600
An arXiv preprint reports results from a 13-task computational benchmark; POPPER recorded higher execution success.
Preprint: Protected pristine devices showed no detectable field-free diode effect, while responses were observed after oxidation and in a selectively etched device.
An arXiv preprint reports that ICON reduced correlation-driven false positives in simulations and exposed shortcut reliance in image and brain scans.
Preprint: An exact analysis finds a lowest band with Chern number -1 and modest, nonzero geometric defects, while the middle pair touches.
Preprint testing reports higher retrieval scores on the PAB anomaly-search benchmark and transfer to a separate person-retrieval dataset, with limits on complex queries.
A preprint reports long-lived coherence among about 100 trillion spins and a billion microwave-cavity photons, while identifying three distinct operating regimes.
Preprint simulations compared two 30-nucleotide strands in magnesium, calcium and spermine; fitted persistence length leveled off above 0.5 millimolar spermine.
Preprint: An LLM-based score aligned more closely with psychometric patterns than BLEU and cosine measures in one item bank, while simulated adaptive tests showed a trade-off.
Preprint: A focused search reports nine improved q-ary covering-code upper bounds for alphabet sizes six and seven, with all nine codes checked independently.
Preprint modeling finds that BOA’s carbon break-even point depends heavily on the electricity mix, storage assumptions and compression performance.
Preprint: Under a regular-sequence assumption, Ext-module membership implies that matching local-cohomology modules belong to the same Serre subcategory.
A Rocq-verified model of inlining offers formal guarantees, but it does not test production compilers, hardware or real compile times.
Preprint tests NEO-sDFT against NEO-DFT in water dimers and measures NEO-FDE scaling in larger clusters, with errors growing at the shortest tested displacement.
Preprint: A mathematical analysis finds when symmetric transport systems have a unique solution that preserves its norm, while rougher coefficients can permit non-uniqueness.
Preprint: Noise-based spectra recovered modeled plasma frequencies across several dimensions, while rotating-wall tests showed nonlinear mode mixing and weak-drive compression responses.
An arXiv preprint tests an adaptive shield in five benchmark environments, with safety, reward and exploration varying by setting.
The August 2026 preprint proposes a laboratory comparison of a conventional code pull request and a reviewable “spec-delta” workflow, but reports no condition-level findings.
Preprint: A proof-based method separates local constraints and derives an exact dimension formula plus lower and upper bounds under stated mesh assumptions.
A preprint under review reports gains over PPO and GRPO, while lower measured runtime was reported only in one ALFWorld comparison.
A conceptual preprint defines resilience as maintaining and restoring service performance in dynamic, uncertain environments, with four dimensions and six proposed metrics.