## Reviewer 1 ### Strengths (1) The authors clearly describe three research questions they want to address, and conduct the experiments with them. The research questions are meaningful and common in the real-world application. (2) The illustrations (diagrams, formula etc.) provided are comprehensible. (3) The proposed method consistently outperforms the baseline across all evaluation tasks. ### Weaknesses (1) The task used for evaluation is somewhat singular. Its empirical study was barely conducted on macroeconomic data from South Korea. This method should be able to applied to other similar issues facing the same problems (as illustrated in the 3 research questions). However, this work did not show this, even did not show whether it could be applied to other regions. Thus, cannot convincingly show its great impact to the real-world. (2) It lacks of description of the selection of those macroeconomic indicators. This is very important in economic research. In this study, I think the report time of those macroeconomic indicators are more critical than the time labels, as some macroeconomic indicators might also have delay in time of publication like the GDP. Nowcasting should guarantee only use those data accessible before the current time. (3) 1 have doubts about the science and applicability of using cubic spline interpolation in the method. Which point, and how many points are involved for the cubic spline interpolation, should be justified. This fitting method is sensitive to the data points. If all data points in train/dev/test set are used for that, there might be an information leakage from the future time step. If only the data points from the previous time step are involved here, it might be better to provide some details on how to process the first few intervals. Because it makes little sense to fit a cubic spline to fewer than five points. **Overall evaluation:** -1 (weak reject) ### Overall evaluation (comments) Based on the Weaknesses (1) and concerns in Weakness (2) and (3), I tend to weak reject this paper. I think this paper could be improved. ## Reviewer 2 ### Strengths 1. The problem studied in this paper is interesting. 2. The proposed method is reasonable. 3. The experimental results seem promising. **Weaknesses:** 1. It would be better if more datasets are used for experiments. **Overall evaluation:** 1 (weak accept) ### Overall evaluation (comments) The paper introduces a new model called NCDENow, which combines Nonlinear Continuous Dynamic Equations (NCDEs) with Dynamic Factor Models (DFMs) to improve the accuracy of nowcasting GDP growth. The proposed method addresses challenges such as sudden drops and irregular or missing data more effectively. The researchers conducted experiments to evaluate the performance of NCDENow in different scenarios, considering factors like missing data rate and volatility of economic data. The results showed that NCDENow outperformed various versions of Recurrent Neural Networks (RNNs) and DFMs in nowcasting GDP. Pros: 1. The problem studied in this paper is interesting. 2. The proposed method is reasonable. 3. The experimental results seem promising. Cons: 1. It would be better if more datasets are used for experiments. ## Reviewer 3 ### Strengths The model is well explained. The experimental setting is clear and the results are promising. ### Weaknesses some minor issues with the choice of the dataset and the lack of additional datasets to compare with. **Overall evaluation:** 1 (weak accept) ### Overall evaluation (comments) In this paper, the authors introduce a novel framework for nowcasting GDP, making the following contributions: - They adopt the framework of Nonlinear Continuous Dynamic Equations (NCDEs) to capture the dynamics of continuous macroeconomic indicators. - They propose a method to address the dynamic factor loading and idiosyncratic disturbance, enabling accurate calculation of GDP growth. The experiments on GDP nowcasting are conducted on the Korean GDP dataset and compared with 6 baselines. As far as my knowledge is concerned, the model seems well explained. The experimental setting is clear and the results are promising. I think that there might be some minor issues with the choice of the dataset and the lack of additional datasets to compare with. - Can you share the dataset you collected from Economics Statistics System? It would be important for both reproducibility and baseline studies. - Independently from the country, could you add in your experiment a dataset already used by other papers, such as https:// onlinelibrary.wiley.com/doi/full/10.1002/jae.2845, to see how your approach works on datasets and methods optimized on that particular dataset? - The improvements shown in Table 2 and Table 3 are very likely significant from a statistical significance point of view. What about the computational complexity of your algorithms compared to the baselines? ## Meta-review ### Meta-review The reviewers found this work interesting and have provided several concrete suggestions to further improve the paper. I encourage the authors to consider the detailed reviewer feedback and address them as appropriate.