Official Review · NEURIPS 2024

Comprehensive Data Integration: CY-Bench combines data from multiple sources, including weather, soil, and remote sensing, providing a rich dataset for model training and evaluation. Standardization and Preprocessing: The standardized preprocessing steps ensure that the data is ready for analysis, reducing the effort required for data preparation. Robust Benchmarking: The benchmark's design allows for robust and reproducible model evaluations, facilitating fair comparisons between different approaches. The authors provide detailed information on data sources, preprocessing steps, and evaluation metrics, which enhances the reproducibility and usability of the dataset. The paper is also written clearly and concisely, making it accessible to readers from various disciplines. The inclusion of tables, figures, and examples helps illustrate the dataset's features and the benchmarking process. The significance of CY-Bench lies in its potential to advance the state-of-the-art in crop yield forecasting by providing a comprehensive and standardized dataset. The benchmark facilitates the comparison of different modeling approaches, helping identify the most effective methods for improving food security and agricultural planning. Pros Comprehensive Dataset: Covers a wide range of countries and crop types, providing extensive data for model training and evaluation. Standardization: Ensures consistent data preprocessing and evaluation, facilitating fair comparisons between different models. Cons No Differentiation Between Irrigated and Non-Irrigated Systems: This could lead to inaccuracies in yield forecasts for regions with mixed farming practices. Absence of Weather Forecast Integration: Models cannot access future weather conditions beyond the lead time, potentially missing critical information that affects crop yields.