Statistical versus machine learning-based spatial interpolation of post-processed ensemble weather forecasts
研究比较了统计和机器学习方法，用于处理德国未观测站点的天气预报插值问题，提出了一种改进的海拔感知线性池方法。
0 citationsRead paper
Selected work
Representative Papers
Extending TotalSegmentator: Predicting Patient and Acquisition Characteristics from CT and MR Images
研究开发并评估了一种快速开源模型，通过CT和MR图像直接预测患者及采集特征，使用3D ResNet-10集成方法训练模型。
0 citationsRead paper
Improving the sharpness in neural network-based parametric post-processing of ensemble forecasts
This study addresses the common trade-off in post-processing ensemble forecasts, where neural network–based calibration often sacrifices predictive sharpness—particularly for short lead times—to improve reliability. To jointly optimize both calibration and sharpness, the authors propose a novel approach that introduces, for the first time, a sharpness penalty term directly into the continuous ranked probability score (CRPS) loss function. Assuming a Gaussian predictive distribution, the method is evaluated on ECMWF 2-meter temperature ensemble forecasts and achieves a reduction of 8.2%–12.5% in the width of central prediction intervals while maintaining CRPS and mean RMSE at levels comparable to baseline methods. This demonstrates a significant enhancement in forecast sharpness without compromising overall forecasting skill.
0 citationsRead paper
AI and physics-based weather forecasting: A comparative study
This study systematically evaluates the accuracy and computational efficiency of AI-based versus physics-based models in medium-range forecasting of 10-meter wind speed. Leveraging observational data from over 9,000 global stations during July–November 2025, it presents the first large-scale operational comparison between ECMWF’s AI-driven AIFS and its traditional physics-based IFS model, complemented by both parametric ensemble model output statistics (EMOS) and nonparametric quantile regression (QR) for post-processing to correct systematic biases. Results show that the raw IFS significantly outperforms AIFS; however, both post-processing methods substantially improve forecast skill—with EMOS yielding greater gains—and markedly reduce the performance gap. After calibration, IFS retains only a marginal advantage at short lead times. This work provides critical empirical evidence supporting the operational deployment of AI-based weather prediction systems.
0 citationsRead paper
Error-Driven Prompt Optimization for Arithmetic Reasoning
In regulated domains such as finance and healthcare, industrial agents must perform arithmetic reasoning over structured data locally while ensuring sensitive information remains within the organizational boundary—posing challenges for accuracy, privacy, interpretability, and parameter efficiency. Method: We propose an error-driven prompt optimization framework that automatically clusters model prediction errors, iteratively refines interpretable prompt rules, and requires no parameter fine-tuning. Integrating error analysis, hierarchical clustering, and dynamic prompt engineering, the method establishes an end-to-end arithmetic reasoning enhancement pipeline on the Qwen3-4B small language model. Contribution/Results: Under strict local deployment constraints, our approach achieves 70.8% arithmetic reasoning accuracy—significantly outperforming GPT-3.5 Turbo—and marks the first solution enabling high-accuracy, privacy-preserving, interpretable, and zero-shot fine-tuned table-based arithmetic reasoning in fully localized settings.
0 citationsRead paper
Recent publications
Latest Papers
Statistical versus machine learning-based spatial interpolation of post-processed ensemble weather forecasts
研究比较了统计和机器学习方法，用于处理德国未观测站点的天气预报插值问题，提出了一种改进的海拔感知线性池方法。
0 citationsRead paper
Extending TotalSegmentator: Predicting Patient and Acquisition Characteristics from CT and MR Images
研究开发并评估了一种快速开源模型，通过CT和MR图像直接预测患者及采集特征，使用3D ResNet-10集成方法训练模型。
0 citationsRead paper
Improving the sharpness in neural network-based parametric post-processing of ensemble forecasts
This study addresses the common trade-off in post-processing ensemble forecasts, where neural network–based calibration often sacrifices predictive sharpness—particularly for short lead times—to improve reliability. To jointly optimize both calibration and sharpness, the authors propose a novel approach that introduces, for the first time, a sharpness penalty term directly into the continuous ranked probability score (CRPS) loss function. Assuming a Gaussian predictive distribution, the method is evaluated on ECMWF 2-meter temperature ensemble forecasts and achieves a reduction of 8.2%–12.5% in the width of central prediction intervals while maintaining CRPS and mean RMSE at levels comparable to baseline methods. This demonstrates a significant enhancement in forecast sharpness without compromising overall forecasting skill.
0 citationsRead paper
AI and physics-based weather forecasting: A comparative study
This study systematically evaluates the accuracy and computational efficiency of AI-based versus physics-based models in medium-range forecasting of 10-meter wind speed. Leveraging observational data from over 9,000 global stations during July–November 2025, it presents the first large-scale operational comparison between ECMWF’s AI-driven AIFS and its traditional physics-based IFS model, complemented by both parametric ensemble model output statistics (EMOS) and nonparametric quantile regression (QR) for post-processing to correct systematic biases. Results show that the raw IFS significantly outperforms AIFS; however, both post-processing methods substantially improve forecast skill—with EMOS yielding greater gains—and markedly reduce the performance gap. After calibration, IFS retains only a marginal advantage at short lead times. This work provides critical empirical evidence supporting the operational deployment of AI-based weather prediction systems.
0 citationsRead paper
Error-Driven Prompt Optimization for Arithmetic Reasoning
In regulated domains such as finance and healthcare, industrial agents must perform arithmetic reasoning over structured data locally while ensuring sensitive information remains within the organizational boundary—posing challenges for accuracy, privacy, interpretability, and parameter efficiency. Method: We propose an error-driven prompt optimization framework that automatically clusters model prediction errors, iteratively refines interpretable prompt rules, and requires no parameter fine-tuning. Integrating error analysis, hierarchical clustering, and dynamic prompt engineering, the method establishes an end-to-end arithmetic reasoning enhancement pipeline on the Qwen3-4B small language model. Contribution/Results: Under strict local deployment constraints, our approach achieves 70.8% arithmetic reasoning accuracy—significantly outperforming GPT-3.5 Turbo—and marks the first solution enabling high-accuracy, privacy-preserving, interpretable, and zero-shot fine-tuned table-based arithmetic reasoning in fully localized settings.
0 citationsRead paper