A ConvLSTM trained on Ceará satellite data from 2000–2023 captured regional land-degradation trends better than changes in individual 2.2-kilometre cells. The researchers say cell-level changes remained largely unpredictable.
Source: arXiv, Desertification Forecasting in Semi-Arid Regions: Using Deep Learning to Probe the Observability Limits of Land Degradation Dynamics (arXiv:2610.13406)
Stanford researchers’ Chronos Policy method aligns robot observations and actions using timestamps rather than fixed steps. The preprint reports better retention of task success under some timing changes.
Source: arXiv, Chronos Policy: Learning Robot Policies on Physical Time (arXiv:2610.13929)
Authors affiliated with CNRS IRISA in Rennes report that, in synthetic network experiments, reverse diffusion recovered a noisy training graph in ways standard sample-level metrics missed.
Researchers at the universities of Oslo and Xi’an presented a graph neural network for probabilistic power-flow estimates when grid topology changes after line outages. The model outperformed baselines on phase-angle error in two test networks.
Source: arXiv, Hierarchical Graph Neural Network for Probabilistic Power Flow with Grid Topology Contingencies (arXiv:2610.13495)
A Michigan Tech study found that LiDAR vehicle-detection accuracy rose with snowfall intensity in two snowy-road datasets. The authors attribute the unexpected association to confounding changes in object point density and geometry.
Source: arXiv, Beyond the Snowflakes: Disentangling Weather Effects from Confounding Meta-Factors in LiDAR Object Detection (arXiv:2610.13891)
A preprint by authors affiliated with Paris Dauphine-PSL and CNRS evaluates the out-of-sample fit of search-based feature selection for explanatory NASDAQ-100 regression.
An arXiv preprint by an author affiliated with the University of Oslo gives constructive bounds linking neural-network architectures to the topology of their representations.
Researchers at Ohio State and Wisconsin–Madison presented sparse amortized variational inference for Gaussian processes with spatial data. In a NOAA Arctic sea-ice example, the model used 1,398 variational parameters for 21,855 observations.
Georgia Tech researchers’ Splat2Act system scans an indoor space by phone, builds a digital twin and adapts a robot’s object-pickup policy with simulated demonstrations. The paper tested task success on a robot arm.
Source: arXiv, Splat2Act: From Captured Scenes to Personalized VLA Manipulation (arXiv:2610.13945)
Researchers in the United States and Canada presented SecFD-FL, a method for forecasting electric-vehicle charging loads without centralizing distributed data. The study tested forecasts and gradient-inversion attacks on Palo Alto station records and EV WATTS data.
Source: arXiv, SecFD-FL: Frequency-Decoupled Federated Learning with Secure Aggregation for Charging Load Forecasting in Internet of Electric Vehicles (arXiv:2610.12909)
Researchers affiliated with NYU Abu Dhabi present a framework for testing whether model-written quantum software tests distinguish correct from faulty solutions.
Researchers affiliated with French institutions presented a robot-failure detector that tracks how objects are positioned and related during a task. They evaluated it on an existing benchmark and a new dataset with two tasks.
Source: arXiv, Robot-WATCHDOG: failure detection through object-centric graph representation (arXiv:2610.13403)
Google Analytics can measure visits and site activity using cookies if you allow it. You can change your choice later.
Measurement and privacy details
Google Analytics measurement starts only if you allow it. After consent, automatic measurement may cover scrolls, outbound link clicks, downloads, video engagement, site searches and basic form interactions; search terms may be detected from URL parameters. This site’s own code does not send form field values; email redaction in automatic measurement is best-effort, not guaranteed. Our manual page-view event uses a cleaned URL. You can change your choice later. Visits that begin with consent may also send numeric LCP, INP and CLS performance measurements after Analytics loads. A first or later grant starts this on the next page load; withdrawing consent stops these performance reports for this page. The preference cookie lasts six months; advertising measurement and personalization stay off.
Your browser is sending a Global Privacy Control signal. Analytics is off by default; you may explicitly change this choice here.
