Researchers have introduced GHGbench, a new benchmark and dataset designed to unify and improve the prediction of carbon emissions at both company and building levels. The benchmark addresses fragmentation in existing datasets by providing a comprehensive collection of company disclosures and harmonized building data across multiple cities. Initial findings highlight that predicting building emissions is more challenging than company emissions, and that generalization to new regions or cities is a significant hurdle, with multimodal remote-sensing embeddings proving particularly useful. AI
Summary written by gemini-2.5-flash-lite from 2 sources. How we write summaries →
IMPACT Provides a unified benchmark to advance AI research in predicting carbon emissions, potentially aiding climate change mitigation efforts.
RANK_REASON The cluster describes a new academic paper introducing a benchmark and dataset for carbon emission prediction.