Subject
Algorithms, software, networks, cybersecurity, automation, and robotics.
A shared humanoid policy reported higher success than LHM-Humanoid on seen and unseen object-transport tasks, but success fell as the number of objects increased.
Scientific publication4 min read
A preprint describes a robot that watches for meaningful changes after human-object interactions and uses a vision-language model to plan assistance without a task request. The reported comparison is described as comparable, but the available extraction omits the numerical results and uncertainty estimates needed to judge how close the systems were.
Scientific publication4 min read
CrashSim generated safety-critical scenarios that tracked real-world crash behavior and crash-type statistics more closely than competing methods. In closed-loop tests, the resulting nuCrash set produced much wider performance differences among five autonomous-driving planners.
Scientific publication3 min read
The tested methods reached more than 99% of the study's HVmax on several graph sets, while the preprint says they may not recover every best trade-off.
Scientific publication4 min read
A review of public discussions, interviews and repositories finds that many software specifications mix requirements with implementation detail, while practitioners use the workflow selectively.
Scientific publication4 min read
A preprint study of software repositories found that coding-agent adoption coincided with more activity, but repositories with more structured committed AI practices had smaller increases in some static-analysis quality measures. The authors stress that the findings are associations, not proof that configuration caused the difference.
Scientific publication4 min read
In a laboratory test with one human collaborator and a seven-degree-of-freedom robot, an adaptive controller recorded a 95% average success rate, compared with 83% for a fixed-stiffness version and 69% for a deterministic baseline.
Scientific publication4 min read
A theoretical preprint finds that delayed cache misses preserve key online bounds, while offline victim choice and unequal fetch costs remain harder.
Scientific publication4 min read
A preprint proposes a 41-part vocabulary for construction activities and demonstrates four selected physical actions on robot hardware.
Scientific publication3 min read
A preprint reports that cached Event Tokens produced the best reported ranking score and higher retrieval scores across several industrial recommendation comparisons.
Scientific publication4 min read
A mathematical preprint gives deterministic bounds linking a code’s excess quadratic discrepancy to how unevenly its correction-radius balls cover a finite Hamming space. The result does not show closeness to any particular perfect code or establish that all admissible perfect codes exist.
Scientific publication4 min read
LM-X recorded higher average task success than the tested controls while exposing three online signals for progress, next events and local action reliability.
Scientific publication3 min read
TacForcing updates unfinished action blocks with execution-time tactile feedback and reported the highest average success rates in both simulated and real-world evaluations. The study was limited to six tasks in each setting and two robot platforms.
Scientific publication4 min read
Six vision-language models showed more agreement when ranking 3D objects along affective, or feeling-related, adjective pairs than along unrelated pairs, but less than on geometric controls. The pattern varied sharply across object categories and changed when researchers altered the text used to define each axis.
Scientific publication5 min read
EVEREST recorded the strongest overall mask-overlap and pixel-level precision-recall scores on SocioSeg, while the study does not show whether the result generalizes beyond that benchmark.
Scientific publication4 min read
A preprint study reports that CRAMER adapted a frozen sequential recommender to natural-language requests, outperforming the strongest baseline in most tested settings while adding less reported inference overhead.
Scientific publication4 min read
A test-time adaptation method for vision-language navigation recorded higher reported scores than the cited baseline on three unseen-environment benchmarks, while its real-world evidence remained qualitative.
Scientific publication4 min read
A multi-signal detector for robot image streams reported close-to-per-frame segmentation accuracy while sharply reducing how often the system adapted.
Scientific publication4 min read
A software gate that compared Python and SPARQL calculations agreed on all seven tests, yet three published counts were wrong.
Scientific publication3 min read
A preprint examines two io_uring vulnerabilities alongside Linux 6.18's SLUB sheaf/barn caching and reports three exploitation techniques, including a configuration-specific local privilege escalation.
Scientific publication4 min read
A robotics preprint reports higher success and shorter solver times when a planner branches as other agents' intentions are revealed, while stressing that the evidence comes from simulations.
Scientific publication4 min read
A small benchmark found that coding agents could turn detailed specifications and empty workspaces into mostly working repositories, but the last few failures often came from narrow, easy-to-miss requirements.
Scientific publication5 min read
A small observational study found that software-development tasks involving generative AI were associated with higher self-rated cognitive load, but task type mattered more and wearable data added little.
Scientific publication4 min read
A failure-bank method reported higher success and lower policy-induced cost than base policies in ten-task averages for π0.5 and π0, but its mean policy cost was higher than AEGIS.
Scientific publication4 min read
A modeling comparison on a reconstructed 2015 Ukraine Power Grid network found that CVSS, EPSS, SSVC and IronMiner can produce different priorities. The authors said the systems capture overlapping but non-equivalent dimensions of vulnerability risk.
Scientific publication4 min read
A new benchmark found that coding agents spent much of their effort locating relevant code across large systems, while the best tested setup completed 26.3% of tasks.
Scientific publication5 min read
A recommender-system method called Mult-BiW reported stronger offline accuracy and larger gains for less-popular items, while falling short of the strongest baseline on two Google metrics.
Scientific publication3 min read
The paper reports higher measured throughput in 37 of 40 GLM 5.2 and MiniMax M2.7 settings and the highest measured throughput in 21 of 28 Kimi K3 settings, while throughput fell in one reuse ablation after the combined MPU phase.
Scientific publication4 min read
A single-host benchmark found that a separate assurance layer could refresh security posture below a five-second software delivery gate, while leaving detection accuracy untested.
Scientific publication4 min read
A preprint presents a symbolic-numeric eigensolver that recovered repeated solutions in a two-equation polynomial example and reconstructed a sample symmetric tensor. The reported results illustrate the method, but do not establish broad speed, stability, or accuracy.
Scientific publication4 min read
An arXiv preprint defines one-, three- and six-component probability functions for LET+K and reports formal equivalence results for its semantic and syntactic updates.
Scientific publication3 min read
A theoretical and computational preprint examines how the Bernard-Letac construction turns repeated draws from a biased source into a uniform outcome. It reports exact fairness, sharper prime-modulus cost results, and favorable results against two procedural comparators in the listed benchmark configurations.
Scientific publication4 min read
A computational framework matched simulations closely in a design case, but its estimates became much less reliable in some low-capacity settings.
Scientific publication5 min read
A theoretical analysis says Khatri-Rao sketches can reach a nearly linear dependence on subspace dimension, narrowing a gap with unstructured random sketches.
Scientific publication4 min read
A preprint reports that BERT had the strongest overall scores for assigning cyber threat intelligence events to sectors, with a caveat about unequal sector coverage.
Scientific publication4 min read
A test-suite-free framework built around an AI model produced C candidates that often compiled and matched held-out input-output checks, but the benchmark also exposed false acceptances and weaker results on stripped binaries.
Scientific publication4 min read
A theoretical preprint offers a direct Haar-based analysis of randomized quasi-Monte Carlo integration and derives an expected squared-error bound under specified function assumptions, but reports no experiments.
Scientific publication4 min read
HubMixer sends recommendation features through compact learned hubs before mixing them and writing them back to individual tokens. The preprint reports the best offline AUC on four objectives, fewer parameters than RankMixer and TokenMixer, and a 5.48% online conversion improvement before full deployment in the tested business.
Scientific publication4 min read
An updated Magma benchmark separates reaching a bug, satisfying its trigger condition, and detecting the resulting fault. In reported 24-hour campaigns, 77 of 127 bugs were reached and 43 were triggered, while AFL++ recorded the highest observed trigger count.
Scientific publication4 min read
An arXiv preprint presents a unified Rényi framework for composite binary testing, combining finite-sample bounds with asymptotic results that separate two Type II error regimes.
Scientific publication4 min read