pipette
ESEspañol

Friday, October 2

2,333 new papers appeared. We read every one. These 20 are worth your time.

Each dot is one paper. The pink ones made today's edition. Point at any dot to see its title.

Today's edition

Picked for broad interest, a real step forward and honest claims, with no field allowed to take over.

  1. Towards Physical Underwater Robotic Assistance for Scuba Diver Movement in Confined Spaces

    We demonstrate that relatively low thrust values (10 percent of maximum) allow robotic direction of a human's movement using the physical sensation of the robot's guidance.
    AI summary

    The authors developed RADMCS, a wearable robot that helps divers stay at a fixed distance from structures using camera-based depth sensing and thruster force feedback. They tested the device with eight divers in pools and ocean water, measuring sensitivity thresholds and distance-keeping ability. The robot can guide a diver's movement using only 10 percent of its maximum thrust, letting the person feel the direction through physical sensation.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big stepReal-world use
  2. Population connective field modeling reveals proto-retinotopic visual cortex organization in the prenatal human brain

    We found topographically organized eccentricity and polar angle connectivity maps in V2 and V3 of the visual cortex as early as the 21st week of gestation.
    Peer-reviewed journalClaims a big step
  3. A highly potent bispecific antibody confers complete protection against SFTSV in preclinical models

    More importantly, in preclinical ferret models, the bsAb N2D2-Fc-N2C1 conferred complete protection against SFTSV not only at a low dose of 5 mg/kg but also when administered as late as 2 d postinfection.
    AI summary

    Severe fever with thrombocytopenia syndrome virus (SFTSV) is an emerging bunyavirus that can kill up to 30% of infected people, and no approved treatments exist. The researchers found two nanobodies, N2C1 and N2D2, that block SFTSV by stopping both viral attachment and membrane fusion. Structural analysis showed N2C1 spans two domains of the viral protein, while N2D2 targets one domain, and together they work synergistically. A bispecific antibody combining both nanobodies showed dramatically enhanced neutralizing potency and superior protection in mice compared to the individual nanobodies. In ferrets, the bispecific antibody N2D2-Fc-N2C1 gave complete protection at just 5 mg/kg, even when given two days after infection.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Peer-reviewed journalClaims a big stepReal-world use
  4. SonarVoxNet: Diver Detection in 3D Bounding Box using 3D Sonar

    First, SonarVoxNet adapts a voxel-based encoder and an anchor-free center-based detection head to 3D sonar data, replacing the conventional yaw-only rotation representation with a continuous 6D rotation parameterization to predict full 9-DoF oriented bounding boxes -- to our knowledge, the first 3D sonar diver detector to do so.
    AI summary

    Autonomous underwater vehicles need to track where divers are and which way their bodies point. Cameras work poorly underwater, and regular sonar loses the up-down angle needed to know body tilt. New 3D sonar keeps this angle but gives sparse, noisy data, and existing tools only handle upright objects. The authors built SonarVoxNet, which adapts existing methods to 3D sonar and tracks divers in any body position. They also created Diver3D, the first public dataset of 3D sonar images showing divers in varied poses with full orientation labels. Tests showed that allowing full rotation instead of just left-right turning greatly improves detection and reduces orientation error. The results suggest that 3D sonar alone can recover full body orientation, enabling future work on diver pose and robot interaction.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big stepReal-world use
  5. A learned oral signal drives long-lasting thirst satiation

    These findings reveal that the experience of rapid thirst quenching is not a hard-wired response to the consumption of water, but rather a learned prediction of future bodily state.
    AI summary

    Drinking quenches thirst before water is absorbed into the body. The researchers used sham drinking to separate water intake from actual rehydration. Sham drinking suppressed thirst and inhibited thirst-promoting neurons without any water absorption. This suppression matched the amount of water consumed and lasted for tens of minutes after drinking stopped. The effect is triggered by water passing through the mouth and is learned, not hard-wired.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Preprint
  6. Long-range transport of biomolecular condensates by Marangoni effect

    Here, combining microfluidic reconstitution, quantitative micropipette aspiration and live-cell imaging, we show that a MEG-3 concentration gradient generates interfacial-tension asymmetry and drives the directed transport of PGL-3-RNA co-condensates in vitro and P granules in vivo through the Marangoni effect.
    AI summary

    They found that a protein gradient creates uneven surface tension that pushes condensates in one direction. This Marangoni effect drives transport both in lab experiments and in living cells. The team could predict condensate movement without adjusting any parameters. Their work shows that physical properties of condensates themselves help organize cell contents.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big step
  7. Methylation-associated Gene Expression (MaGE): a non-invasive, in vivo measure of cellular senescence.

    We developed Methylation-associated Gene Expression (MaGE) predictors: whole-blood DNA methylation proxies of p14ARF, p16INK4A and p21CIP1 expression, plus a composite score, providing the first scalable measure of senescence-marker expression.
    AI summary

    They tested it in 18,859 people and found 45 significant links to 18 different diseases and overall death. MaGE-p21 linked to liver disease from alcohol, while MaGE-p16 linked to lung scarring, matching known patterns of where these proteins act. In another study, MaGE tracked how severe Crohn's disease was and how well treatment worked. The authors say MaGE allows large-scale aging research and could help select patients for trials of drugs that clear senescent cells.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big stepReal-world use
  8. Decoding the Disaster: Multi-Task Geospatial Reasoning with Vision-Language Models and Crowdsourced Imagery for Disaster Mapping

    To our knowledge, this study provides the first systematic investigation and unified evaluation framework for examining how VLM-based spatial reasoning can transform crowdsourced disaster imagery into actionable geospatial artificial intelligence (GeoAI) through cross-view geolocalization validation, interpretable spatial reasoning, and damage-aware severity assessment.
    AI summary

    The authors built a system called GRDisaster that uses vision-language models to understand and locate crowdsourced disaster photos. They created a dataset of 26,340 images from 2018 to 2024, organized into triplets of volunteer photos, street-view images, and satellite images. The system matches volunteer photos to known locations by combining street-view and satellite images. It uses structural and environmental cues to check location matches and assess visible damage with expert-verified labels. This is the first systematic study of how vision-language models can turn crowdsourced disaster images into actionable geospatial intelligence.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big stepReal-world use
  9. Amplitude-controllable event-driven organic photosensors based on ionic-mediated inhibition

    Here we report an event-driven organic photosensor that integrates temporal-contrast detection and amplitude modulation within a single active layer by coupling ionic and electronic transport in an organic mixed ionic–electronic conductor.
    AI summary

    The authors built an event-driven organic photosensor that detects changes in light and adjusts signal strength in one layer. Fast electron extraction creates excitatory spikes, while ion-compensated hole accumulation provides voltage-tunable control over spike amplitude. This mimics attentional gain modulation in biological vision. The sensor preserves motion-relevant contrast and reduces redundant read-out. The authors establish ionic-electronic coupling as a materials strategy for adaptive, low-power neuromorphic vision hardware.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Peer-reviewed journalReal-world use
  10. Independent Amplitude and Phase Modulation in Active Transmissive Metasurfaces via Complex Permittivity Tuning

    Here, we theoretically demonstrate an active transmissive metasurface that enables independent amplitude and phase modulation in the mid-infrared.
    AI summary

    Their device uses germanium cube-shaped resonators combined with graphene strips. Electrical gating of graphene changes light transmission from nearly 0% to 5% at one frequency and shifts phase by 281 degrees at another frequency. They demonstrated beam steering with over 90% efficiency by controlling each unit cell individually.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big stepReal-world use
  11. Analogue black hole lightring and its resonances in an optical fiber

    We demonstrate the complete observation of a black-hole lightring resonance in an optical analogue.
    AI summary

    The authors observed a black-hole lightring resonance in an optical analogue system. They extended the photon sphere concept to non-stationary systems by linking the lightring to a spatiotemporal caustic. For an optical soliton acting as a black-hole analogue, they predicted a spectral plateau with boundaries set by the real part of the lightring frequency. Using a pump-probe experiment in a 5-km-long fiber, they observed these signatures for two pump wavelengths. The measured spectra matched simulations and the predicted lightring-mode profile, giving direct access to the resonance's oscillatory and decay properties.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big step
  12. Single-cell profiling and genetic regulation of alternative polyadenylation in the human brain

    Here we report a single-cell atlas of APA of the aged human brain across 2 million cells from 379 human postmortem brains across individuals with and without Alzheimer’s disease (AD).
    AI summary

    The authors created a single-cell atlas of alternative polyadenylation (APA) in 2 million cells from 379 aged human brains, including people with and without Alzheimer's disease. APA changes in Alzheimer's disease involve different genes than expression changes, but both affect similar pathways like microglial activation and neuron transport. They found cell-type-specific genetic variants affecting APA for 4,288 genes, which more often affect protein levels than gene expression levels. Of 168 disease-related genetic loci linked to these APA variants, only 17.5% overlap with expression-related loci, including variants for Alzheimer's, schizophrenia, and Parkinson's disease. For Parkinson's disease gene SNCA, the APA variant changes where the mRNA sits in the cell, acting independently of expression changes.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Peer-reviewed journalClaims a big step
  13. Stressed alga-free coral cells can initiate coral bleaching

    These findings identify a host-initiated, intercellularly propagating route to coral bleaching and reveal an early host response that can determine the subsequent fate of the symbiosis.
    AI summary

    Scientists used brief stress on corals to study when bleaching begins. They found coral cells without algae turned positive for reactive oxygen species first. This early response spread between cells through the surrounding liquid. ROS-positive alga-free cells alone could trigger bleaching in unstressed corals. Blocking these early cells reduced later bleaching more than later intervention.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Preprint
  14. Chiral trions in enantiopure single-wall carbon nanotubes

    In this study, the CD of trions in enantiopure single-wall carbon nanotubes (SWCNTs) is demonstrated.
    AI summary

    Scientists had not observed trion circular dichroism caused purely by chiral structure. This study shows trion circular dichroism in pure single-wall carbon nanotubes. The sign of this effect flips with right- or left-handed nanotubes. Chiral trions exist in chiral carbon nanotubes.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    Peer-reviewed journalClaims a big step
  15. Near-Noiseless Single-Photon Detection in the Infrared with a Megapixel Semiconductor Array for Low-Background Astronomy

    Here we demonstrate single-photon-resolving operation in a megapixel-format HgCdTe linear-mode avalanche photodiode array operated at conventional cryogenic temperature.
    AI summary

    Sensor noise blocks clear detection of single photons in faint astronomical imaging. Infrared astronomy usually needs superconducting detectors that demand extreme cold and high power. The authors built a megapixel array of mercury-cadmium-telluride avalanche photodiodes working at standard cryogenic temperatures. Avalanche gain lifts photon signals above noise while barely amplifying dark signals, yielding about 2 false counts per thousand reads per pixel. These linear-mode avalanche photodiodes may enable future infrared photon-counting cameras for low-background telescopes.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintClaims a big stepReal-world use
  16. ARCCS: An Automated Regulatory Compliance Checking System

    We present ARCCS, an end-to-end, automated, agentic, and regulation-agnostic Legal NLP system for compliance checking.
    AI summary

    ARCCS is a computer system that automatically checks whether documents follow legal rules. It breaks down complex regulations into simple, traceable requirements and evaluates documents using evidence and confidence scores. The system works with any regulation, regardless of its size or structure. In GDPR policy tests, judges found its decisions legally consistent in up to 96.67% of cases. On an EU procurement benchmark with over 1,200 rule checks, it achieved 98.8% accuracy in detecting violations.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintReal-world use
  17. An auditable multimodal agentic system for sequential decision-making across the acute stroke pathway

    StrokeAgent demonstrates the feasibility of a decision support system that integrates native multimodal data processing with auditable, evidence supported recommendations aligned with the sequential nature of acute stroke care.
    AI summary

    It reads patient records, lab results, heart rhythm tracings, and brain scans to give step-by-step recommendations. In tests with 100 complex stroke cases, StrokeAgent agreed with expert decisions 86% of the time. This was 13 points better than standard AI tools and 17 points better than practicing neurologists given the same information.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintReal-world use
  18. Science Utopia? Closed-Loop LLM Simulation of Academic Research Ecosystems

    Using these longitudinal simulations, we find that rejection-driven resubmission substantially amplifies reviewer burden beyond population growth alone, cautious exploration balances citation impact with career success and long-term topic diversity, and resource inequality can emerge even without detectable cumulative advantage from narrowly winning early funding.
    AI summary

    The authors built SciUtopia, a computer simulation that uses AI agents to mimic how academic research works over time. The simulation includes researchers, institutions, funding, collaborations, peer review, and how scientists choose what to study. They ran 61 simulated worlds with over 40,000 researchers from 8,000 institutions, producing about 400,000 publication decisions and 1.2 million AI-generated reviews. They found that resubmitting rejected papers greatly increases reviewer workload beyond what population growth alone would cause. Careful exploration of new topics balances citation impact with career success, and unfair resource distribution can arise even without clear early advantages.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintCode available
  19. Detecting Inconsistencies in Model Specifications with LLM-as-Verifier Reasoning

    We introduce VeriSpec, the first approach to directly detect inconsistencies in model specifications by auditing the specification text itself.
    AI summary

    Model specifications guide how large language models should behave, but two principles may conflict in the same situation. The authors created VeriSpec, which uses an LLM as a verifier to find inconsistencies directly in natural-language specification text. Applied to the OpenAI Model Spec, VeriSpec extracted 405 rules and found five validated inconsistencies, which were reported to developers. VeriSpec achieved 38.5% precision and cost $11.12 per validated inconsistency, outperforming five baselines. The authors suggest direct specification auditing can catch defects before they shape any model behavior.

    Written by Kimi K2.5 from the abstract only. Jev checked every sentence against the authors' words. See where each sentence comes from

    PreprintReal-world useCode available
  20. Self-organized hummock-hollow patterns redistribute salinity stress in coastal meadows

    Together, these results reveal that spatial self-organization can support ecosystem functioning and resilience under climate stress through an underappreciated mechanism: the spatial redistribution of environmental stress.
    Peer-reviewed journalReal-world use