Multi-document summarization (MDS) aims to compress the content in large\ndocument collections into short summaries and has important applications in\nstory clustering for newsfeeds, presentation of search results, and timeline\ngeneration. However, there is a lack of datasets that realistically address\nsuch use cases at a scale large enough for training supervised models for this\ntask. This work presents a new dataset for MDS that is large both in the total\nnumber of document clusters and in the size of individual clusters. We build\nthis dataset by leveraging the Wikipedia Current Events Portal (WCEP), which\nprovides concise and neutral human-written summaries of news events, with links\nto external source articles. We also automatically extend these source articles\nby looking for related articles in the Common Crawl archive. We provide a\nquantitative analysis of the dataset and empirical results for several\nstate-of-the-art MDS techniques.\n