LaDe: The First Comprehensive Last-mile Express Dataset from Industry

1. Although the proposed dataset can support the research in several research topics, it can only support a narrow research area in AI. Thus, its impacts may be limited.

2. The proposed dataset is more suitable for data mining tasks. It may not be very relevant to this conference, which focuses on more general deep learning methods. It should be better to release this dataset in a data mining conference/journal.

1. The limited geographic coverage of this dataset in five Chinese cities raises concerns about its broader relevance and transferability to different geographic and cultural contexts.
2. Another limitation could be missing timestamp and location of all the trajectories covered by couriers, currently, only pickup and drop locations and corresponding time are recorded.

1. The paper is not in the scope of ICLR. ICLR is a top-tier venue for sharing the development of machine learning and its applications. The released dataset has a limited contribution to the machine learning community. Although the authors have discussed three applications, they are not the main topic of ICLR.

2. The dataset is collected based on the real-world service on the logistics platform, and thus it is not general. The authors have not considered an important question: what about other logistics platforms with different express services? Will the dataset be useful?

3. I suggest the authors propose novel machine learning methodologies for machine learning tasks, including classification, prediction, etc., for this dataset.

W1. The data fields of the dataset proposed in the article appear relatively straightforward and, in essence, do not significantly distinguish themselves from the data fields found in other existing Pick-up & Delivery datasets.

W2. The dataset outlined in this research paper reveals certain limitations with respect to its incorporation of road network structure, a pivotal aspect for applications pertaining to pick-up and delivery services. Predominantly relying on GPS locations(x-y coordinates), while undoubtedly valuable, might not comprehensively address the intricacies inherent in real-world logistics operations. The absence of road network information in the dataset does raise valid concerns about its suitability for applications necessitating an in-depth understanding of spatial constraints and the optimization of routing, particularly in the last-mile express domain. To augment the dataset's applicability in such scenarios, it may prove advantageous to contemplate the inclusion of coarse-grained road network data. While acknowledging the privacy concerns associated with road network information, providing a more generalized form of road network data could still facilitate researchers in achieving a more comprehensive and accurate representation of the real-world scenarios under examination.

W3. The dataset should acknowledge the significant impact of rush hours and traffic congestion on logistics. It is important to recognize that the road network structure undergoes dynamic and temporal changes over the course of several months. Therefore, it becomes paramount to incorporate a dynamic road network structure within the dataset that can adapt to and depict the wide-ranging traffic conditions and road configurations at different times. This dynamic approach not only bolsters the dataset's precision and relevance but also furnishes researchers with a more authentic representation of the continually evolving real-world logistics and last-mile delivery scenarios.

W4. It is worth noting that several datasets, including prominent examples like the Didi Chuxing GAIA dataset and the New York City dataset, have been extensively employed in addressing pick-up and delivery problems. These datasets have formed the foundation of numerous benchmarks and research benchmarks, rendering them well-established within the field. Consequently, it is essential to delineate the distinctive advantages that the dataset introduced in this paper brings to the table in comparison to these existing datasets and benchmarks, particularly in the context of last-mile delivery research. Clarifying the unique attributes and contributions of this new dataset is crucial in highlighting its relevance and significance within the academic and research community. To accentuate the contributions of the article, it is essential to delve deeper into a comprehensive comparison that delineates the differences between the dataset tailored for Last Mile Delivery scenarios and datasets associated with broader Pick-up & Delivery contexts, including well-known examples like the DIDI and NYC taxi datasets.

W5. Throughout the article, the authors seek to benchmark several aspects of applications, each of which boasts a significant body of existing works. Yet, the chosen methods in the article appear to fall short of encompassing the full spectrum of relevant works and categories within these benchmarked areas. In light of this, it is imperative that the authors offer more extensive elucidation regarding their decision to select these specific methods as baselines when there are numerous alternatives at their disposal. Providing a comprehensive explanation will not only enhance the transparency and academic rigor of their research but also enable the reader to grasp the underlying reasoning behind these choices.