Preprint automates geospatial forecasting from data selection through model training

Published Monday, September 21, 2026 · 5:09 PM EDT

Researchers introduced the Planetary Prediction Engine, a preprint system that turns natural-language questions into geospatial forecasts by finding relevant open-web and Earth-observation data, combining it with foundation-model map embeddings and selecting a task-specific model with overfitting checks. The authors report that the system raised mean R² across 21 U.S. health indicators from 60.0% to 76.8% and improved Nigerian food-security prediction from 31.5% to 66.1%. For weekly forecasts of the 2026 Democratic Republic of Congo Bundibugyo Ebola outbreak, it identified 15 of 18 newly affected health zones. These results are preprint claims across selected tasks and have not established performance in routine operational use.

Sources

cs.AI updates on arXiv.org

Published by Tech & Business, a media brand covering technology and business. This story was sourced from cs.AI updates on arXiv.org and reviewed by the T&B editorial agent team.

Back to Newswire

Keep reading

Full wire

Science AI

Bloomberg preprint reports 729 successful migrations using AI-assisted pipelines

Bloomberg researchers report 729 successful code migrations from 2,333 pull requests generated during two production campaigns, in a new preprint.

Science AI

Preprint reports AI search improves quantum error-correction codes

Researchers report in an arXiv preprint that a large language model found quantum error-correction codes that improve on published benchmarks in selected parameter regimes.

Science AI

EviSeek preprint reports higher accuracy with fewer tokens for embodied agents

Researchers report that EviSeek, a framework requiring no additional training, improved agents' accuracy on questions that require gathering visual evidence through movement and physical interaction.

Science AI

FuseAMC preprint reports better radio signal classification on one CPU core

Researchers report in a preprint that FuseAMC, a model for identifying how radio signals encode information, reached 55.8% overall accuracy on RadioML 2016.10a while training and running on one CPU core.

AI Capital

Numica raises $21M to acquire small US accounting practices

Numica has raised $21M in equity and debt from Roosh X and Flashpoint Capital to acquire more small US accounting firms, The Next Web reports. The New York startup targets practices with annual revenue of $700,000 to $5m, often owned by accountants approaching retirement.

Capital AI

Kabilio raises €10M to develop AI agents for accounting firms

Barcelona startup Kabilio has raised €10M in a round led by Picus Capital and Visionaries, with participation from Spain's technology and innovation agency CDTI. The financing brings its total funding to €14m and will fund an AI agent for accounting firms.