Science in plain language, with the paper one click away Science in plain language
In a five-practice Australian pilot, the software could not read doctors' free-text notes, where 98 percent of the relevant detail lived, and nearly all the flagged patients were being followed up anyway.
In 52 healthy adults learning a joystick task, people carrying lower-dopamine gene variants improved less when told they were beating others than when they got plain timing feedback; the effect was small, exploratory, and needs replication.
In a study of 32 Italian university students with dyslexia and 60 without, both groups picked up the rules of an invented symbol grammar, and the dyslexic readers leaned harder on frequency patterns to compensate.
A new arXiv preprint from a University of North Carolina statistics group argues that patient-simulating models fail in two specific ways, and it is a conceptual argument rather than an experiment with data.
Across 1,213 outpatient visits in May 2023, only 14 percent involved conditions plausibly tied to ash, and daily particle levels showed no significant link to illness, though the design cannot track what any individual actually breathed.
A preprint compiling 432 host-parasite records from 90 sources, plus new sampling in five seas, finds only larval worms in gelatinous animals, suggesting they ferry parasites up the food chain rather than ending their journey.
A re-analysis of 69 archived black garden ant trajectories, plus 60 new ants tested under white and red light, found their walks are biased toward the release point, even though the original study concluded the ants had no sense of position.
New computer models of stars between 1.4 and 2.5 solar masses suggest the peculiar Am stars and their ordinary-looking cousins are the same objects caught at different ages, though the work is a theoretical calculation rather than a set of observations.
A preprint modelling neutron stars laced with "mirror" dark matter finds both vibration frequencies follow tidy patterns that barely depend on the poorly known physics of dense nuclear matter, though the authors caution the cleaner of the two relations weakens once the star's warped spacetime is allowed to wobble too.
A four-page CHI workshop paper claims the emotional moments in hard thinking are part of the thinking itself, and that today's AI tools for thought treat them as either noise or a dial to tune.
A team probing the internal vector spaces of causal, masked, and diffusion language models found that lopsided relations like "part of" carve out clearer territory than mutual ones like synonymy.
Three researchers build a mathematical account of what happens when many writers lean on the same language model, and find that personalized models preserve variety while shared ones pull everyone toward a single norm.
Among 1,502 commercially insured Americans who had part or all of their stomach removed for cancer, the 15 percent carrying a documented depression or anxiety diagnosis stayed longer, had more complications, and cost more to care for, though the study cannot say why.
Chinese chemists built a photoanode that converts cyclohexanone to adipic acid with about 95 percent selectivity, then wired it into a flow battery that produces the acid and supplies current at the same time.
Italian researchers built a browser-based system that walks clinicians through radiomics analysis of brain scans while exposing the intermediate files most AI tools keep hidden.
In a preregistered experiment, an AI that supplied starting ideas narrowed the range of what a whole group of writers produced, while an AI that only polished writers' own ideas left that range intact.
Researchers at Deezer tested whether the digital fingerprints of AI-generated audio survive being mixed with human recordings, and report early success at spotting which individual part came from a machine.
Engineers at Marvell report that swapping electrical switches for a passive fiber shuffle gave 16 host machines a shared 32-terabyte memory pool, more than halving access latency versus conventional CXL pooling in their emulation.
In a new arXiv position paper, Chen Shani maps out how concept structure gets into large language models and argues the field has been recovering it after the fact when it could be designing it in from the start.
A Japanese team tested a gel made from Mongolian Lycium ruthenicum on lab-grown skin cells and two mouse models of atopic dermatitis, and found it calmed inflammation, tightened the skin barrier, and quieted itch-sensing nerves.